Started in January,1974(Monthly)
Supervised and Sponsored by Chongqing Southwest Information Co., Ltd.
ISSN 1002-137X
CN 50-1075/TP
CODEN JKIEBK
Editors
Current Issue
Volume 46 Issue 11, 15 November 2019
  
Surveys
Survey on Smart Contract Based on Blockchain System
FAN Ji-li, LI Xiao-hua, NIE Tie-zheng, YU Ge
Computer Science. 2019, 46 (11): 1-10.  doi:10.11896/jsjkx.190300013
Abstract PDF(1481KB) ( 1217 )   
References | Related Articles | Metrics
Blockchain is a decentralized global distributed database leger.Smart contract is a piece of event-driven program with states that runs over blockchain systems,which can take custody over digital assets.Smart contracts running on a common platform can also implement parts of the functions of traditional applications.Development of the blockchain provides an appropriate platform for smart contract,and smart contract plays an important role on blockchain systems.With the rapid development of blockchain platforms such as Bitcoin and Ethereum,smart contracts have a good development opportunity.However,applications of smart contract are still in the early stage of development,and there are relatively few related studies.The application scenarios of smart contracts are not enough in practical application.This paper studied programming languages and implementation technologies of smart contract,discussed and explored the development status as well as challenges and future prospects.It described the characteristics of different development languages and took a comparison among them.Then,it classified blockchain systems based on the running environment of smart contract,and studied the development,deployment and running mechanism of smart contracts in various blockchain systems.Also,this paper explored the application scope of various smart contract platforms,and took a comprehensive comparison of different blockchain systems on smart contract development,community support and corresponding ecosystems.It introduced the status and challenges of smart contract research,and conducted analysis on security,scalability,and maintainability.Finally,it analyzed the development trend of blockchain and smart contract technology,and discussed the application scenarios in the future.
Research Advances and Future Challenges of FPGA-based High Performance Computing
JIA Xun, QIAN Lei, WU Gui-ming, WU Dong, XIE Xiang-hui
Computer Science. 2019, 46 (11): 11-19.  doi:10.11896/jsjkx.191100500C
Abstract PDF(1580KB) ( 797 )   
References | Related Articles | Metrics
Improving the energy efficiency and satisfying the performance need of emerging applications are two important challenges faced by current supercomputing systems.Featured with low power consumption and flexible reconfigurability,FPGA is a promising computation platform for overcoming the above challenges.To explore the feasibility,performance of high-performance computing (HPC) kernels on FPGA has been analyzed by extensive researches.How-ever,kernel of convolutional neural network is not considered in these studies,and the analysis lacks a high-performance processor for reference.Aiming at the dominant kernels in today’s HPC landscape,including breadth-first search,sparse matrix vector multiplication,stencil,smith-waterman and convolutional neural network,this paper summarized the implementation and performance optimization of these kernels on FPGA.Meanwhile,a comparison between FPGA and SW26010 many-core processor regarding their performance and energy efficiency was conducted.Furthermore,major problems of adopting FPGA for constructing HPC systems were also discussed.For the kernels considered in this paper,FPGA can outperform SW26010 processor by 63x in terms of energy efficiency.As for performance of emerging applications like graph analytics and deep learning,FPGA can outperform SW26010 by 26x.Lower communication overhead,better programmability and more integral software library for scientific computing will make FPGA an amenable platform for future supercomputing systems.
Survey of Concepts Related to Data Assets
YE Ya-zhen, LIU Guo-hua, ZHU Yang-yong
Computer Science. 2019, 46 (11): 20-24.  doi:10.11896/jsjkx.190800019
Abstract PDF(1228KB) ( 487 )   
References | Related Articles | Metrics
Under different technological,social and economic backgrounds,different terminologies such as Information assets,Digital assets and Data assets were created due to people’s different understanding about contents in cyberspace.Because the term Asset is related to Resource,Capital and Economy,a series of concepts are extended,such as Information Resources(Capital,Economy),Digital Resources(Capital,Economy),Data Resources(Capital,Economy),etc.This paper reviewed these concepts.Based on the physical attributes,existence attributes and information attributes of data in Big Data context,this paper proposed and advocated the standardization of these concepts into Data Resources,Data Assets,Data Capital and Data Economy,which will be helpful to the exploitation of data resources.
Survey on DNA-computing Based Methods of Computation Tree Logic Model Checking
HAN Ying-jie, ZHOU Qing-lei, ZHU Wei-jun
Computer Science. 2019, 46 (11): 25-31.  doi:10.11896/jsjkx.181102091
Abstract PDF(1421KB) ( 330 )   
References | Related Articles | Metrics
Computation tree logic (CTL) model checking is an important approach to ensuring the correctness and reliability of systems.However,the severe spatio-temporal complexity problems restrict the application of CTL model checkingin industry.The large-scale parallelism of DNA computing and the huge storage density of DNA molecules provide new ideas for resolving the problems.The background and the principle of DNA-computing based methods of CTL modelchecking were introduced.The research progress was reviewed from three aspects:the improvement of power,the improvement of autonomy and the resolution of related problems.Firstly,the research progress of methods in terms of power was summarized from checking only one basic CTL-formula to general CTL-formulas,from CTL-formulas with only future operators to CTL-formulas with past-time operators,and from CTL-formulas to linear temporal logic,projection temporal logic and interval temporal logic formulas.Secondly,the research progress of methods in terms of autonomy from non-autonomous methods based on manual operations of memory-less filtering models to autonomous methods based on molecular autonomy of sticker automata was reviewed,showing that the methods are highly autonomous.At last,relevant problems in improving the predictive efficiency of specific hybridization of DNA molecules and constructing DNA molecules of CTL-formulas were described.In the end,corresponding research directions were discussed by concerning different methods,new models and new applications.
Survey of Research on Computation Unloading Strategy in Mobile Edge Computing
DONG Si-qi, LI Hai-long, QU Yu-ben, ZHANG Zhao, HU Lei
Computer Science. 2019, 46 (11): 32-40.  doi:10.11896/jsjkx.181001872
Abstract PDF(1756KB) ( 891 )   
References | Related Articles | Metrics
Improvement of technology makes smart mobile devices more and more popular.Mobile device traffic is growing rapidly.However,due to the limited resources and computing performance of smart mobile devices,mobile device may face the situation of insufficient capacity when dealing with compute-intensive and time-sensitive applications.Unloading the computations which the mobile terminal needs to process to the computing nodes in the edge network for calculation is an effective way to solve this problem.This paper first introduced the existing calculation offloading strate-gies and elaborated from the aspects of minimizing delay,minimizing energy consumption and maximizing benefits.Then,it compared the advantages and disadvantages of different offloading strategies.At last,it considered and prospected the future development of calculation offloading strategies of mobile edge network.
Research Progress on Data Query Technology in Dynamic Wireless Sensor Networks
LIANG Jun-bin, MA Fang-qiang, JIANG Chan
Computer Science. 2019, 46 (11): 41-48.  doi:10.11896/jsjkx.181202258
Abstract PDF(1513KB) ( 131 )   
References | Related Articles | Metrics
Wireless sensor networks (WSN) are a self-organizing network composed of a large number of sensor nodes with limited communication,computing and storage capabilities,which can be deployed to perform long-term monitoring tasks in harsh environments.Data query processing is one of the most basic operations for WSN to obtain monitoring data.It mainly refers to the user distributing query requests to the network through a specific node,and then the node that satisfies the requirements returns the data to the user.In the process of query,because of the dynamic nature of the network (e.g.,the destruction of nodes by external forces,the movement or sleep,resulting in changes in network topology and connectivity,and unreliable communication links,etc.),the data transmission has large delay,high energy consumption and even data loss,resulting in low success rate of query.At present,many scholars study this problem and make some progress,but there are still many problems to be solved in practical application.In order to further promote the in-depth study of data query technology in dynamic wireless sensor networks,this paper analysed and summarized the typical work in recent years,and compared their advantages and disadvantages.Then,this paper discussed the key issues that need to be solved in this field,and finally pointed out the next research directions.
Weakly Supervised Learning-based Object Detection:A Survey
ZHOU Xiao-long, CHEN Xiao-jia, CHEN Sheng-yong, LEI Bang-jun
Computer Science. 2019, 46 (11): 49-57.  doi:10.11896/jsjkx.181001899
Abstract PDF(2223KB) ( 1316 )   
References | Related Articles | Metrics
Object detection is one of the fundamental problems in the field of computer vision.Currently,supervised learning-based object detection algorithm is one of the mainstream algorithms for object detection.In the existing researches,high-precision image labels are the precondition of supervised object detection to gain good performance.How-ever,it becomes more and more difficult to gain accurate labels due to the complexity of background and variety of objects in a real scenario.With the development of deep learning,how to receive good performance with the low-cast image labels becomes the key point in this field.This paper mainly introduced object detection algorithms based on weakly supervised learning with image-level labels.Firstly,this paper described the background of object detection and pointed out the shortcomings of training data.Then,it reviewed weakly supervised object detection algorithm based on image-level labels from three aspects:image segmentation,multi-instance learning and convolutional neural network.The multi-instance learning and convolutional neural network were comprehensively illustrated in several ways like saliency learning and collaborative learning.Finally,this paper compared mainstream algorithms based on weakly supervised learning horizontally and compared them with object detection algorithms based on supervised learning.The results prove that weakly supervised object detection algorithm has achieved great progress,especially the convolutional neural network has greatly promoted the development and gradually replaced multi-instance learning.After taking fusion algorithm,its accuracy rate is remarkably increased to 79.3% on Pascal VOC 2007.However,it still performs worse than supervised object detection algorithm.To achieve better performance,the fusion algorithm based on convolutional neural network is becoming a mainstream algorithm in weakly supervised object detection.
Network & Communication
Time Slot Optimization for Channel Hopping in CRN
JI Yi, JIA Jun-cheng, SHENG Kai
Computer Science. 2019, 46 (11): 58-64.  doi:10.11896/jsjkx.181001865
Abstract PDF(2611KB) ( 127 )   
References | Related Articles | Metrics
With the rapid development of wireless communication technology in recent years,radio spectrum resources are becoming scarcer.Cognitive radio networks (CRNs) attracts widespread attention because they can improve the utilization of existing spectrum resources.For the issue that the traditional random channel hopping rendezvous strategy of cognitive radio network do not considere the channel collision and switching delay,this paper proposed an optimal random channel hopping rendezvous strategy based on time slot ALOHA protocol with calculation of switching delay.Firstly,the proposed strategy differentiates the whole process in time slot,defines the model about time slot of channel staying and switching delay,integrates the channel hopping process with ALOHA protocol,and gives the calculation formula of time-to-rendezvous (TTR).Then,by analyzing the process of rendezvous strategy step by step,it derives the formula of time slot expectation for available channel number and switching delay based on joint probability.Finally,it calculates the lowest point according to the derivative and the trend graph of the function.And then according to the idea of integer programming,this paper proposed an algorithm for calculating the optimal number of slots to optimize the overall rendezvous strategy.The experiment was carried out under the control of the number of available slots and switching delay.The experimental results show that the effect of switching delay on the rendezvous efficiency is greater than that of channel number.Also,the proposed scheme can achieve rendezvous in an optimal way with full account of the switching delay.It also can reduce the total rendezvous time effectively than the traditional strategy.When the number of time slots of switching delay is not more than 5,the ATTR is generally reduced by about 15%,which can promote the rendezvous and accelerate the message exchange between nodes and futher improve the spectrum utilization.
3D Node Localization Algorithm Based on Iterative Computation for Wireless Sensor Network
JIANG Rui, WU Qian, XU You-yun
Computer Science. 2019, 46 (11): 65-71.  doi:10.11896/jsjkx.181001855
Abstract PDF(1796KB) ( 157 )   
References | Related Articles | Metrics
The existing three-dimension localization algorithm for wireless sensor networks (WSN) is mostly based on the principle of two-dimension localization algorithm with mature and good performance.Compared with the two-dimensional localization algorithm,most of the three-dimension localization algorithms have better localization accuracy.Two-dimensional localization algorithm for wireless sensor networks based on centroid iteration estimation reduces the range of two-dimensional plane of unknown nodes and improves the positioning accuracy of nodes by iterating the centroid of the plane surrounded by connected anchor nodes.Based on the theory of the two-dimension centroid localization algorithm,this paper proposed a novel approach of three-dimensional node localization algorithm based on iterative computation.First,the centroid coordinates of the three-dimensional space enclosed by the current connected anchor nodes as well as the received signal strength (RSSI) between the unknown node and the centroid node are calculated.Then,the current connected anchor node with the weakest RSSI is replaced with the centroid node in order to reduce the three-dimensional space enclosed by the connected anchor nodes.The location accuracy can be improved through multi-itera-tions with an appropriate threshold.Simulation results are obtained with interactive data language (IDL) on a PC of 3.50GHz.Judging from the simulation results,there is an improvement of 3% to 6% in location accuracy compared with the two-dimension localization algorithm,and an improvement of 5% to 23% in location accuracy compared with the 3D centroid localization algorithm.What’s more,the proposed algorithm performs well for RSSI error disturbance and can reach more than 99% of localization coverage after multi-iterations.
Adaptive Information Transmission Scheme for LEO Satellite Based on Ka Band
YU Xiu-lan, WANG Si-yi
Computer Science. 2019, 46 (11): 72-79.  doi:10.11896/jsjkx.181001862
Abstract PDF(4241KB) ( 222 )   
References | Related Articles | Metrics
Aiming at solving the problem that Ka-band satellite communication signal transmission is easily affected by rainfall and ground environment,this paper studied the distribution of the Ka-band based rain attenuation of LEO satellite channel by considering the high-speed movement characteristics of LEO satellite and the propagation characteristics of satellite-to-earth link,and proposed an adaptive information transmission scheme for LEO satellite to improve the reliability of system information transmission.Firstly,a Ka frequency band LEO satellite mobile communication channel model is established to solve the problem that the transmission signal is easily affected by rainfall and the variation of elevation angle.Secondly,the probability density function of the rain attenuation based on the change of the elevation angle is calculated by combining the elevation varying range and the elevation probability density function of LEO satellite.Thirdly,according to the probability density function of the rain attenuation and the current channel state,the channel parameters are calculated and the channel state information can be obtained.Then,according to the confirmed channel state information,the signal-to-noise ratio thresholds of different channel states and different modulation coding modes can be calculated with the error rate of 1×10-4.Finally,by comparison of the feedback signal-to-noise ratio and the calculated signal-to-noise ratio thresholds,the best modulation and coding mode is selected through the adaptive modulation coding selection algorithm,which can promote the transmission reliability of spatial information.The simulation results show that the lower the elevation angle of satellite is,the more serious the ground shadowing is,and the bit error ratio of system is higher under the different channel states.The proposed adaptive information transmission scheme makes the bit error ratio of the satellite communication systems always lower than the target bit error ratio of 1×10-4.The results show that the proposed adaptive information transmission scheme can effectively solve the problem of serious attenuation of transmission signal caused by rainfall,ground mobile environment and satellite mobility,and effectively enhance the information transmission quality of LEO satellite system based on Ka band.
Bit Error Rate Analysis of Diffusion-based Multicast Molecular Communication Networks
CHENG Zhen, ZHAO Hui-ting, ZHANG Yi-ming, LIN Fei
Computer Science. 2019, 46 (11): 80-87.  doi:10.11896/jsjkx.181001925
Abstract PDF(1727KB) ( 82 )   
References | Related Articles | Metrics
Considering a multicast molecular communication network consisting of one transmitter nanomachine,two receiver nanomachines,and four nanomachines acting as relays,this paper proposed two different relay schemes using the same type and different types of molecules in each hop to transmit information to ensure the reliability of the multicast molecular communication.First,the method for adjusting the decision threshold as an effective mechanism is proposed to mitigate interference when transmitting the same type of molecules between parallel relay nanomachines.Then,mathematical expressions for the average bit error rate of the multicast molecular communication network for both relay schemes are derived.Finally,the simulation results show that different parameters have impacts on the average bit error rate of the multicast molecular communication network,including decision threshold,the number of molecules emitted in each time slot,the distance between transmitter nanomachine and receiver nanomachine,the number of samples,bit interval duration and diffusion coefficient.And a relay scheme which can reduce the average bit error rate of this network was proposed.
Dynamic Resource Allocation for UAV Video Uploading
HE Chao, XIE Zhi-dong, TIAN Chang
Computer Science. 2019, 46 (11): 88-93.  doi:10.11896/jsjkx.190500106
Abstract PDF(2178KB) ( 149 )   
References | Related Articles | Metrics
Unmanned Aerial Vehicles (UAV) can capture images and videos in real time by the sensors they carry.In particular,when a cluster of UAVs work together,they are able to efficiently complete reconnaissance,perception,mapping and other tasks,which make them widely used in both military and civil fields.However,all videos captured by UAVs need to be transmitted to the ground station or control center through wireless channels.The requirement of wireless channel transmission rate is higher and higher,along with video service definition unceasing enhancement and the cluster quantity continuous increase.Thus,under the constraint of limited wireless transmission resources,how to allocate them in the UAV cluster to maximize the uploading quality of videos has become an urgent problem to be solved.For this problem,a distributed algorithm was designed.In order to specify video from other data transmission,the QoE-oriented utility function is considered first.Then,around the problem,a potential game model is formulated and all the users can update their strategies with very little information exchange.The algorithm converges to a set of correlated equilibria and achieves the global optimal allocation of wireless resources in the cluster.This algorithm starts from the perspective of video application,and according to the properties of different video signals,each UAV can intelligently adjust the channel resource occupation.The highest total utility of the UAV cluster can be achieved under the limited wireless channel resources.Numeric simulation results indicate that it brings remarkable benefits to both resource providers and UAV video users.
Intelligent Incentive Mechanism for Fog Computing-based Multimedia Systems with Swarming Behavior
LIU Lu, ZHAO Guo-qing
Computer Science. 2019, 46 (11): 94-99.  doi:10.11896/jsjkx.181001975
Abstract PDF(2027KB) ( 123 )   
References | Related Articles | Metrics
In order to improve the efficiency of multimedia data transmission and system execution and reduce the ope-rating cost of multimedia services,this paper proposed an intelligent incentive mechanism based on fog computing from the analysis model and evolution of group behavior of multimedia systems.Firstly,based on the characteristics of simplification,decentralized deployment,redundancy,robustness and self-management,an evolutionary model of group behavior analysis for distributed multimedia systems is established,and an evolutionary algorithm of group behavior analysis for multimedia systems is presented.Then,in order to maximize the system utility,the fog server nodes are scheduled by self-organization and active evolution.In order to optimize individual service strategy,fog computing combines with evolutionary process to control group behavior participation.On this basis,the fog server nodes update individual sche-duling step by step,and real-time statistics the system topology scheduling effect.The simulation experiment is based on the simulation platform of networked control system of Matlab,and the multimedia system is deployed.The topology and wireless transmission of distributed multimedia system are simulated by Matlab.The EMSSB (Evolution algorithm of Multimedia Systems Swarming Behavior) algorithm and IIFS (Intelligent Incentive algorithm with Fog computing and Swarming Behavior) algorithm proposed above are implemented in combination with C language.The data of simulation experiments are the average of 100 repetition delays. In each repetitive experiment,the other parameters are consistent except that the time and number of multimedia requests are set to random.The simulation results show that the proposed incentive algorithm performs well in real-time multimedia data transmission,fog node incentive effectiveness and user request response.The proposed incentive algorithm can shorten the end-to-end delay by 45%,effectively control the participation degree,and control the different participation proportion according to the user’s request.In addition,the user response delay and multimedia data stream transmission delay can be reduced by 53% and 45%,respectively.
Information Security
Method of System Safety Analysis and Verification for SysML Models
LI Wan-qian, HU Jun, CHEN Song, ZHANG Wei-jun
Computer Science. 2019, 46 (11): 100-108.  doi:10.11896/jsjkx.181001850
Abstract PDF(2171KB) ( 131 )   
References | Related Articles | Metrics
In recent years,with the increasing scale and complexity of safety-critical systems such as aviation,transportation and medical treatment,model-based system safety analysis and verification has become an important research direction in the field of safety-critical system engineering.How to analyze and verify the safety of SysML as a typical system model is a very important issue.Based on the framework of model-based safety analysis,this paper designed a safety analysis and verification method for the SysML models,which realizes the complete process from model construction to safety analysis and attribute verification.Firstly,this paper introduced SysML system architecture design model and the latest system safety modeling language AltaRica3.0 from requirement level and design level,constructed the semantic equivalence transformation rules from the core model elements of SysML to AltaRica 3.0,gave the formal description of transformation rules,analyzed and proved the transformation rules.Then,this paper designed a prototype tool platform based on the model-driven method to complete the process of automatic transformation and safety analysis of the model.The prototype tool combines the functions of transformation,compilation and generation of fault tree,fault tree analysis,stepwise simulation and dynamic demonstration of fault path.The synchronization of system design and safety analy-sis was realized.On the basis of this,the key points of transformation from AltaRica 3.0 to Promela model were given,and the attributes of the model were verified by the exhaustive model verification tool SPIN.Finally,the SysML model was established according to the architecture design description and safety requirements of the wheel brake system in 4761 standard,and the automatic conversion,safety analysis and verification of the model were realized based on the prototype tool platform and attribute verification tool,which further illustrates the effectiveness of the conversion method.
QoS Quantification Method for Web Server with Mimic Construction
ZHANG Jie-xin, PANG Jian-min, ZHANG Zheng, TAI Ming, LIU Hao
Computer Science. 2019, 46 (11): 109-118.  doi:10.11896/jsjkx.181001922
Abstract PDF(1781KB) ( 84 )   
References | Related Articles | Metrics
As the emerging “Internet Plus” has quickly become an important driving force of social and economic deve-lopment,Web service plays an increasing role in society,but its security issues are worsening.The Web server with mimic construction is a new Web defense system based on the principle of mimic defense,and it uses the heterogeneity,dynamics,redundancy and other characteristics to block or disrupt network attacks.Although it has been deployed and some better defense effects have been gotten,there is still a lack of effective methods for quantifying its QoS.On the basis of analyzing the system architecture of the Web server with mimic construction,this paper discussed the difference and issues between the quantification of its QoS and the quantification of traditional Web servers’ QoS,and analyzed the factors affecting its QoS.Based on the “Wood Barrel” principle,this paper proposed a quantitative evaluation method for the service quality of the Web server with mimic construction,and used the vector similarity method to quantify the loss value of the QoS.This effort provides a new method for quantifying the QoS of the Web server with mimic construction in theory,and provides guidance for optimizing its service quality in engineering practice.The simulation and experimental results show the proposed quantification method can effectively quantify and evaluate the QoS of the Web server with mimic construction compared with the existing evaluation methods.
Improved Higher-order Meet-in-the-Middle Attack on Camellia-256
ZHANG Li, WEI Hong-ru
Computer Science. 2019, 46 (11): 119-122.  doi:10.11896/jsjkx.180901786
Abstract PDF(1346KB) ( 96 )   
References | Related Articles | Metrics
Camellia is an iterated block cipher with Feistel structure.Theblock length of Camellia is 128bits,and the key length is 128bits,192bits or 256bits,which employs a total of 18 rounds for a 128-bit key and 24 rounds for a 192-bit or 256-bit key.At present,the security analysis of Camellia is a research hotspot.According to the key schedule and relation,this paper analyzed the relation between the round keys and found 8 relations of the guessing keys in total by means of the key-bridge technology.Therefore,when 16 rounds Camellia-256 against higher-order meet-in-the-middle attack,the number of subkeys required to compute the relevant values is reduced.The time complexity is reduced by 28.This result is better than any previously published cryptanalytic results on Camellia without FL/FL-1 functions and whitening layers.
Research on Broker Based Multicloud Access Control Model
ZHAO Peng, WU Li-fa, HONG Zheng
Computer Science. 2019, 46 (11): 123-129.  doi:10.11896/jsjkx.190300112
Abstract PDF(1901KB) ( 117 )   
References | Related Articles | Metrics
Multicloud is increasingly accepted by industry and has great promotional value and development potential,since it combines cloud resources in a provider-independent way and there is no need to change the provider’s original technology solutions and operation model.Cloud broker provides transparent service for providers and users,composes the resource of cloud providers on demand,and reduces the difficulty of Multicloud collaboration,the risk of vendor lock-in and the cost of cloud user.However,the loss of trust and the heterogeneity of access control policy among cloud providers can easily cause security problems,such as privacy leakage and data loss,and affect the promotion and application of Multicloud seriously.Based on the factors,such as trust,context and SLA,Multicloud access control model (MC-ABAC) was proposed.Firstly,the framework of MC-ABAC is constructed to collaborate in Multicloud environments,which consists of Virtual Resource Manager (VRM),Access Control Manager (ACM) and Cloud Access Control Broker (CACB).Secondly,MC-ABAC is designed to achieve trust measurement of cloud providers and authorization management in Multicloud.This model defines subject,resource,environment and operation,and formalizes trust,context,SLA and authorization.Thirdly,the workflow of MC-ABAC is designed to access the resource of multicloud from local provider and CACB respectively.Finally,the simulation environment of MC-ABAC is built by using CloudSim 4.0 and OpenAZ,and used to verify the availability,such as the success rate and the response time of the request.The results show the request success rate of MC-ABAC is about 18% higher than that of ABAC,and whose average response time is better than that of ABAC,when MC-ABAC is used normally and the number of requests is large.
Hierarchical Control Strategy for Data Querying Based on Differential Privacy
LI Sen-you, JI Xin-sheng, YOU Wei, ZHAO Xing
Computer Science. 2019, 46 (11): 130-136.  doi:10.11896/jsjkx.180901690
Abstract PDF(2027KB) ( 123 )   
References | Related Articles | Metrics
Protecting users’ private data is critical in the process of data querying,publishing and sharing.Most of the existing privacy protection models provide a uniform level of privacy protection for all query users of the dataset without considering the different query results of different trust levels.This “one size fits all” approach ignores the differences of the privacy protection requirements between individuals.Multiple query users may have different query privilege and reputation value,and the data privacy attributes of the queries are also different.Therefore,those methods of providing a uniform level of privacy protection cannot meet the differentiated needs of privacy protection.This paper proposed a hie-rarchical query control strategy based on differential privacy.When the query user submits a query request,this method can protect data privacy by adding Laplace noise with different distribution characteristics into the returned results for different trust levels queries.The trust levels are based on the query security trust degree according to the privilege,repu-tation value of users and data privacy attribute.In order to ensure high availability data cannot be obtained by low-level query users,the availability evaluation module is introduced to analyze the data availability while protecting privacy.The simulation experimental results demonstrate that the proposed control model can provide protected data with error rates ranging from 0.1% to 30% for different levels of query users,releasing the important limitation of differential privacy providing only a uniform level of privacy protection,and solving the privacy leakage problem of data query of multi-trust level users.And analyzing the availability of the query results can maximize data availability within the context of diffe-rential privacy protection.
Software & Database Technology
Empirical Study of Code Query Technique Based on Constraint Solving on StackOverflow
CHEN Zheng-zhao, JIANG Ren-he, PAN Min-xue, ZHANG Tian, LI Xuan-dong
Computer Science. 2019, 46 (11): 137-144.  doi:10.11896/jsjkx.191100501C
Abstract PDF(1464KB) ( 121 )   
References | Related Articles | Metrics
Code query plays an important role in code reuse,and the Q&A about code on StackOverflow which is a professionalquestion-and-answer site for programmers is a typical scenario for code reuse.In practice,the manual way is adopted to answer questions,which usually has the disadvantages of poor real-time,incorrect description of problems,and low availability of answers.If the process of code query and search can be automated and replace manual answering, it will save a lot of manpower and time cost.Now there are already many code query technologies,but most lack experie-nce of application in the real case.Based on the ideas of Satsy,this paper implemented the code query technology based on constraint solving for Java language,and designed the empirical study.This paper used StackOverflow as the research object,and mainly studied how to apply the code query technology based on constraint solving of Q&A about code on the website.First of all,the problems on the website are analyzed,and 35 problems with high trafficin Java language are extracted as query problems.Then,about 30000 lines of code are captured from GitHub,and they are converted into the form of constraints as well as built as a large code base to support code query.Finally,through the analysis of the query results of these 35 questions,the practical application effect of the technology on StackOverflow was evalua-ted.The results show that the proposed technology has good practical application effect on the specific questions and code scale studied,and can replace the manual answer on a considerable scale.
Method of Microservice System Debugging Based on Log Visualization Analysis
LI Wen-hai, PENG Xin, DING DAN, XIANG Qi-lin, GUO Xiao-feng, ZHOU Xiang, ZHAO Wen-yun
Computer Science. 2019, 46 (11): 145-155.  doi:10.11896/jsjkx.181102210
Abstract PDF(2474KB) ( 282 )   
References | Related Articles | Metrics
In the era of cloud computing,more and more enterprises are adopting microservice architecture for software development or traditional monolithic application transformation.However,microservice system has high complexity and dynamism.When microservice system fails,there is currently no method or tool that can effectively support the location of the root cause of failure.To this end,the paper first proposed that all business log generated on all of the ser-vices by a single request can be associated by the trace information.And on this basis,this paper studied the method of microservice system debugging based on log visualization analysis.Firstly,the model of microservice log is defined.So the data information required for log visualization analysis can be specified.Then five kinds of visual debug strategies are summarized to support the location of four kinds of typical microservice fault’s root cause.The four kinds of microservice faults are ordinary fault with exceptions,logical fault with no exceptions,fault caused by unexpected service asynchronous invocation sequences and faults caused by service multi-instances.The strategies include single trace with log information,comparison of different traces,service asynchronous invocation analysis,service multi-instances analysis and trace segmentation.Among them,in order to realize service asynchronous invocation analysis and service multi-instances analysis,this paper designed two algorithms.At the same time,a prototype tool named LogVisualization was designed and implemented.LogVisualization can collect log information,trace data,nodes information and service instance information of the cluster,generated by the microservice system runtime.It can associate the business log with trace information by less code intrusion.And it supports users to use five strategies for visual debug.Finally,the prototype tool is applied to the actual micro-service system.Compared with the existing tools (Zipkin+ELK),the usefulness and effectiveness of prototype tool in the root location of four micro-service faults are verified.
Cost-sensitive Convolutional Neural Network Model for Software Defect Prediction
QIU Shao-jian, CAIZi-yi, LU Lu
Computer Science. 2019, 46 (11): 156-160.  doi:10.11896/jsjkx.191100502C
Abstract PDF(1522KB) ( 130 )   
References | Related Articles | Metrics
Machine-learning-based software defect prediction methods are received widely attention from the researchers in the field of software engineering.The defect distribution in the software can be analyzed by the defectprediction mo-del,so as to help the software quality assurance team to detect potential software errors and allocate test resources reasonably.However,most of the existing defect prediction methods are based on hand-crafted features such as line of code,dependency between modules and stack reference depth.These methods do not take into account the potential semantic features of the software source code and may result in poor predictions.To solve the above problems,this paper applied convolutional neural networks to mine the semantic features implicit in the source code.In the effective mining of source code semantic features,this paper used three-layer convolutional neural network to extract data abstract features.In terms of data imbalance processing,this paper adopted a cost-sensitive method,which gives different weights to positive and negative examples,and balances the impact of positive and negative examples on model training.In terms of experimental data sets,this paper selected multiple versions of the eight softwares in the PROMISE defect dataset,totaling 19 projects.In terms of model comparison,this paper compared the proposed cost-sensitive software defect prediction model based on convolutional neural network (CS-TCNN) with logistic regression and deep confidence network respectively.The evaluation metrics contain AUC and MCC,which are widely used in the field of defect prediction research.The experimental results demonstrate that CS-TCNN can effectively extract the semantic features in the program code,and improve the prediction effect of the software defect prediction model.
Stochastic TBFL Approach Based on Calibration Factor
WANG Zhen-zhen, LIU Jia
Computer Science. 2019, 46 (11): 161-167.  doi:10.11896/jsjkx.191100503C
Abstract PDF(1406KB) ( 70 )   
References | Related Articles | Metrics
Approaches for fault localization based on test suites are now collectively called TBFL (Testing Based Fault Localization).However,current algorithms have not taken advantages of the prior knowledge about test cases and program,so that they waste these valuable “resources”.Literature [12] introduced a new kind of stochastic TBFL approach whose spirit is to combine the prior knowledge with actual testing activities under stochastic theory,so as to locate program faults.This algorithm may be regarded as a general pattern of this kind of approach,from which people can deve-lop various algorithms.The approach presented in this paper was simplifying the TBFL algorithm.It mainly revises the prior probability of program variable X from separate testing activity of each test case.If there are n test cases,n calibration factors can be obtained.These n calibration factors are then added and standardized,finally the posterior probability of the program is obtained.The approach proposed in this paper is called stochastic TBFL approach just because it depends on a calibration factor matrix.This paper presented three standards for comparing different TBFL approaches.Based on these standards,the improved approach is feasible for some instances.
Modified Neural Language Model and Its Application in Code Suggestion
ZHANG Xian, BEN Ke-rong
Computer Science. 2019, 46 (11): 168-175.  doi:10.11896/jsjkx.191100504C
Abstract PDF(1928KB) ( 71 )   
References | Related Articles | Metrics
Language models are designed to characterize the occurrence probabilities of text segments.As a class of important model in the field of natural language processing,it has been widely used in different software analysis tasks in recent years.To enhance the learning ability for code features,this paper proposed a modified recurrent neural network language model,called CodeNLM.By analyzing the source code sequences represented in embedding form,the model can capture rules in codes and realize the estimation of the joint probability distribution of the sequences.Considering that the existing models only learn the code data and the information is not fully utilized,this paper proposed an additional information guidance strategy,which can improve the ability of characterizing the code rules through the assistance of non-code information.Aiming at the characteristics of language modeling task,alayer-by-layer incremental nodes setting strategy is proposed,which can optimize the network structure and improve the effectiveness of information transmission.In the verification experiments,for 9 Java projects with 2.03M lines of code,the perplexity index of CodeNLM is obviously better than the contrast n-gram class models and neural language models.In the code suggestion task,the average accuracy (MRR index) of the proposed model is 3.4%~24.4% higher than the contrast methods.The experimental results show that except possessing a strong long-distance information learning capability,CodeNLM can effectively model programming language and perform code suggestion well.
Ensemble Model for Software Defect Prediction
HU Meng-yuan, HUANG Hong-yun, DING Zuo-hua
Computer Science. 2019, 46 (11): 176-180.  doi:10.11896/jsjkx.180901685
Abstract PDF(1263KB) ( 102 )   
References | Related Articles | Metrics
Software defect prediction aims to identify defective modules effectively.Traditional classifiers have good predictive effect on class-balanced data,but when the proportion of data classes is unbalanced,the traditional classifiers incline to majority classes,easily leading to the misclassification of minorityclass module.In reality,the data in software defect prediction are often unbalanced.In order to deal with this kind of class imbalance problem in software defects,this paper proposed an integrated model based on improved class weight self-adaptation,soft voting and threshold mo-ving.This model considers the class imbalance problem in the training stage and decision stage without changing the original data sets.Firstly,in class weight learning stage,the optimal weights of different classes are obtained through class weight adaptive learning.Then,in the training stage,three base classifiers are trained by using the optimal weights obtained in the previous step,and the three base classifiers are combined by soft ensemble method.Finally,in the decision stage,the decision is made according to the threshold moving model to get the final prediction category.In order to prove the validity of the proposed method,the NASA software defect standard data sets and the Eclipse software defect standard data sets are used for prediction,and the proposed method is compared with the results of several software defect prediction methods proposed in recent years on the recall rate Pd,false positive rate Pf and F1 measurement F-measure.The experimental results show that the recall rate Pd and F1 measurement F-measure of the proposed method improves by 0.09 and 0.06 on average respectively.Therefore,the overall performance of proposed method for dealing with class imbalance in software defect prediction is superior to other software defect prediction methods,and it has better prediction effect.
Artificial Intelligence
Multi-modal Sentiment Analysis with Context-augmented LSTM
LIU Qi-yuan, ZHANG Dong, WU Liang-qing, LI Shou-shan
Computer Science. 2019, 46 (11): 181-185.  doi:10.11896/jsjkx.181001941
Abstract PDF(1836KB) ( 478 )   
References | Related Articles | Metrics
In recent years,multi-modal sentiment analysis has become an increasingly popular research area,which extends traditional text-based sentiment analysis to a multi-modal level that combines text,images and sound.Multi-modal sentiment analysis usually requires the acquisition of independent information within a single modality and interactive information between different modalities.In order to use the context information of language expression in each modality to obtain these two kinds of information,a multi-modal sentiment analysis approach based on context-augmented LSTM was proposed.Specifically,each modality is encoded in combination with the context feature using LSTM which aims to capture the independent information within single modality firstly.Subsequently,the independent information of multi-modality is merged,and the other LSTM layer is utilized to obtain the interactive information between the different modalities to form a multi-modal feature representation.Finally,the max-pooling strategy is used to reduce the dimension of the multi-modal representation,which will be fed to the sentiment classifier.The method achieves 75.3% ACC on the MOSI data set and F1 reaches 74.9.Compared to traditional machine learning methods such as SVM,ACC is 8.1% higher and F1 is 7.3 higher.Compared with the current advanced deep learning method,it is 0.9% higher on ACC and 1.3 higher on F1.At the same time,the trainable parameters are reduced by about 20 times,and the training speed is increased by 10 times.The experimental results demonstrate that the performance of the proposed approach significantly outperforms the competitive multi-modal sentiment classification baselines.
DC-BiGRU_CNN Model for Short-text Classification
ZHENG Cheng, XUE Man-yi, HONG Tong-tong, SONG Fei-bao
Computer Science. 2019, 46 (11): 186-192.  doi:10.11896/jsjkx.180901702
Abstract PDF(1877KB) ( 308 )   
References | Related Articles | Metrics
Text classification is a basic task in natural language processing.Nowadays,it is more and more popular to use deep learning technology to deal with text classification tasks.When processing text sequences,convolutional neural networks can extract local features,and recurrent neural networks can extract global features,all of which show good effect.However,convolutional neural networks can not capture the context-related semantic information of text very well,and recurrent networks are not sensitive to the key semantic information.In addition,although deeper networks can better extract features,they are prone to gradient disappearance or gradient explosion.To solve these problems,this paper proposed a hybrid model based on densely connected gated recurrent unit convolutional networks (DC-BiGRU_CNN).Firstly,a standard convolutional neural network is used to train the character-level word vector,and then the character-level word vector is spliced with the word-level word vector to form the network input layer.Inspired by the densely connected convolutional network,a proposed densely connected bidirectional gated recurrent unit is used in the stage of high-level semantic modeling of text,which can alleviate the defect of gradient disappearance or gradient explosion and enhance the transfer between features of each layer,thus achieving feature reuse.Next,the convolution and pooling operation are conducted for the deep high-level semantic representation to obtain the final semantic feature representation,which is then input to the softmax layer to complete text classification task.The experimental results on several public datasets show that DC-BiGRU_CNN has a significant performance improvement in terms of the accuracy for text classification tasks.In addition,this paper analyzed the effect of different components of the model onperfor-mance improvement,and studied the effect of parameters such as the maximum length of sentence,the number of layers ofthe network and the size of the convolution kernel on the model.
Robust SVM Based on Zeroth Order Variance Reduction
LU Shu-xia, CAI Lian-xiang, ZHANG Luo-huan
Computer Science. 2019, 46 (11): 193-201.  doi:10.11896/jsjkx.181001840
Abstract PDF(5002KB) ( 138 )   
References | Related Articles | Metrics
Great losses will be produced when traditional SVM methods are used to deal with the classification problem with noisy data,which makes the classification hyperplane seriously deviates from the optimal hyperplane,resulting in poor classification performance.In order to solve this problem,this paper proposed a robust support vector machine (RSVM) and gave a loss function in sinusoidal square form.According to the characteristics of sinusoidal function,the value of loss function is limited to the range of [0,1],even for noise data,which improves the anti-noise ability of SVM.In addition,when the traditional stochastic gradient descent method is used to solve the SVM,a single sample gradient is used to approximately replace the full gradient in each iteration,which will inevitably produce variance.As the number of iterations increases,the variance also accumulates,which seriously affects the classification performance of the algorithm.In order to reduce the influence of variance,this paper introduced a zeroth order-stochastic variance reduced gradient (ZO-SVRG) algorithm.This algorithm uses coordinate gradient estimation method to replace gradient approximately,and reduces the influence of variance by introducing the gradient correction term in each iteration.Besides,in the output of the internal and external loop,the weighted average output form is adopted,and then the convergence speed of the optimization problem is accelerated.The experimental results show that the robust support vector machine based on zeroth-order variance reduction algorithm has better robustness to noise data and effectively reduces the influence of variance.In order to further improve the performance of the algorithm,the influence of the main parameters λ and k on the accuracy of algorithm were analyzed.For both linear and nonlinear cases,when its parameter pairs (λ,k) are satisfied (λ=1,k=5) and (λ=10,k=3),respectively,the highest accuracy of each can be achieved.
Microblogging Water Army Identification Based on Semi-supervised Collaborative Training Algorithm
HAN Qing-qing, ZHANG Yan-mei, NIU Wa
Computer Science. 2019, 46 (11): 202-208.  doi:10.11896/jsjkx.180901617
Abstract PDF(2258KB) ( 204 )   
References | Related Articles | Metrics
In the fast-developing Internet era,Weibo brings a large amount of information,but there exists water army in Weibo topic.To a certain extent,the water army affects ordinary users to understand the real situation.In order to efficiently and accurately identify the water army,the semi-supervised collaborative training algorithm is considered comprehensively in view of the small number of water military samples and the large number of non-water military samples.By studying and analyzing multiple characteristics of Weibousers,the proposed algorithm redefines six attribute feature values,such as account attention,daily microblog number,and microblog influence.According to the characteristics of the algorithm,the six attribute feature values are divided into two attribute sets,each attribute set corresponds to one view,and each view uses seven classification methods in the Scikit-Learn machine learning library to train the classifier to identify the water army.Finally,experiments are conducted on dataset.The results show that the accuracy,recall rate,accuracy and F1-measure value of the classification results are higher when the two views use the naive Bayes algorithm and the logistic regression algorithm to train the classifier.Therefore,comprehensive analysis of Weibo user cha-racteristics and the use of semi-supervised collaborative training algorithms in line with the actual situation can accurately,efficiently and quickly identify Weibo water army.
Knowledge Reasoning Method Based on Unstructured Text-enhanced Association Rules
LI Zhi-xing, REN Shi-ya, WANG Hua-ming, SHEN Ke
Computer Science. 2019, 46 (11): 209-215.  doi:10.11896/jsjkx.181001939
Abstract PDF(1906KB) ( 121 )   
References | Related Articles | Metrics
Knowledge bases (KBs) store entities,entity attributes and relations between entities in a structured manner.Because the knowledge in the KBs can be easily processed by computers,KBs play a vital role in many natural language processing (NLP) tasks.Although current KBs contain massive triple knowledge from the perspective of absolute quantity,they are far less than the knowledge existing in real world.Therefore,many researches focus on how to enrich the knowledge base with more high-quality knowledge.Internal reasoning and extracting from external resources are two main kinds of methods for knowledge base completion,but they still need to be improved.On the one hand,since the knowledge in KBs are not perfect and some errors exist,reasoning on such error knowledge will cause error propagation.On the other hand,existing extracting methods usually focus on limited relations and properties and thus cannot find comprehensive knowledge from external resources such as texts.In light of this,this paper proposed a knowledge reasoning method based on unstructured text-enhanced association rules.In this method,the text representation pattern is abstracted from the unstructured text firstly,then it is represented in the form of a bag of distribution,and the associa-tion rules can be mined through combining the knowledge of KBs.The difference from the traditional association rules is that the association rules obtained by the proposed method can directly match unstructured texts for knowledge reasoning.Experimental results show that the proposed method can efficiently infer triple knowledge from unstructured text with higher quality and larger quantity compared with traditional methods.
Influence Space Based Robust Fast Search and Density Peak Clustering Algorithm
CHEN Chun-tao, CHEN You-guang
Computer Science. 2019, 46 (11): 216-221.  doi:10.11896/jsjkx.181001846
Abstract PDF(1481KB) ( 97 )   
References | Related Articles | Metrics
DP (Clustering by Fast Search and Find of Density Peaks) is a novel clustering algorithm based on local density and distance between data,which has the advantages of being able to discover the clusters of arbitrary shapes,easy to accept logic principle and efficiently partitioning data into data sets.However,this algorithm has some disadvantages,for instance,it can’t deal with the case when multiple density peaks co-exist in a single cluster,and it is characterized with the unstable class label allocation.At the same time,it can’t accurately identify the clusters of sparse data when the density difference between clusters is big enough.In order to overcome the above deficiencies,this paper proposed a robust density peak clustering algorithm based on influence space.The proposed algorithm is capable of calculating local density by neighboring data and enhances the ability to recognize small-scale clusters.In order to improve the robustness of data partition,the data influence space is introduced,a new symmetric relationship is defined,and a new allocation strategy is proposed.By calculating the weighted density ratio of the target data to the adjacent data,the algorithm is able to process data sets with multiple density distribution features.The local density is obtained by calculating the density ratio of the target data to the neighbor data,and the ratio is weighted through using the size of the influence space.In order to verify the availability of the proposed algorithm,simulation experiments were conducted on synthetic data sets and UCI data sets,and NMI and Acc indexs were used to evaluate clustering algorithms.Simulation experiment results show that the proposed algorithm can reduce the influence of the cutoff distance on the algorithm and classify the data more stably.Compared with other algorithms,the proposed algorithm achieves better experimental results on both NMI and Acc indexs.
Bus Travel Time Prediction Algorithm Based on Multi-line Information Fusion
MA Lin-hong, CHEN Ting-wei, HAO Ming, ZHANG Lei
Computer Science. 2019, 46 (11): 222-227.  doi:10.11896/jsjkx.180901764
Abstract PDF(1995KB) ( 248 )   
References | Related Articles | Metrics
Aiming at the problems of bus travel time prediction,such as sparse data,lack of data and long update interval,this paper proposed a Kalman filter algorithm based on similar section segmentation and fusion of multi-line information.In this method,the attribute features and spatial structure features of each road segment are normalized,the similar road segments are dynamically divided by using the similarity between the attribute features and the spatial structure and the change of the traffic impact of the POI.Then,the data information of multiple bus lines on similar road segments and target road segments are integrated,and the experimental data are enriched by using the data from similar road segments.Finally,combining the dynamic and real-time characteristics of Kalman filtering algorithm,the model is established to realize short-term prediction and correct the information.In the experiment,162 lines and 299 lines in Shenyang City were selected as experimental lines,and a similar section was taken for basic data collection and experiments.The information on the similar road sections is used to infer sparse information or missing road sections,thereby shortening the data update interval and improving the real-time performance and accuracy of the algorithm prediction.Especially in the early peak period,the absolute average percentage error of the proposed model reaches 13.2%,which can effectively meet the performance requirements of real-time query.
New Neural Network Method for Solving Nonsmooth Pseudoconvex Optimization Problems
YU Xin, MA Chong, HU Yue, WU Ling-zhen, WANG Yan-lin
Computer Science. 2019, 46 (11): 228-234.  doi:10.11896/jsjkx.181001926
Abstract PDF(1474KB) ( 81 )   
References | Related Articles | Metrics
The research of optimization problem is favored by researchers.As an important part of optimization pro-blem,convex optimization problem is the focus of research.Many models based on neural network are applied to practical problems.However,the optimization problems involved in machine learning,signal processing,bioinformatics and other fields are often not convex optimization problems,but pseudoconvex optimization and nonconvex optimization problems.Therefore,it is urgent to solve the latter kind of problems.To solve the optimization problem that the objective function is nonsmooth pseudoconvex function and constraint function is equality and inequality function,this paper constructeda new single-layer neural network model without penalty parameter based on the idea of penalty function and differential inclusion.The main idea of the design is that according to the proposed neural network model,a constrained function can be designed for the gradient of the objective function so that the value of the objective function is always kept within a range,and then a function about time is combined to ensure that its value decreases with time.At the same time,considering that inequality constraints affect the convergence direction of the state solution before it enters the equation constraint,a conditional function is added to restrict it.Compared with the proposed neural network model,it has the advantages of simple structure,no need to calculate penalty parameters in advance,and no special requirements for the position of the initial point.Furthermore,it is theoretically proved that for any initial point,the state solution can converge to the equality constraints in finite time and stay there thereafter,the boundedness of the state solution,the state solution can converge to the feasible region in finite time and stay there thereafter,and the state solution can finally converge to the optimal solution of the optimization problem.Under the environment of MATLAB,by mathematical simulation experiments,the state solution can converge to the optimal solution quickly.At the same time,if the penalty parameters or initial points are not selected properly,the state solution will not converge well when the same optimization problem is solved by using the proposed similar neural network model.This not only verifies the correctness of the theoretical results,but also shows that the proposed network has a wider range of applications.
Questions Recommendation Based on Collaborative Filtering and Cognitive Diagnosis
QI Bin, ZOU Hong-xia, WANG Yu, LI Ji-xing
Computer Science. 2019, 46 (11): 235-240.  doi:10.11896/jsjkx.180901827
Abstract PDF(1451KB) ( 267 )   
References | Related Articles | Metrics
The question recommendation method isthe new application of data mining on the Education Measurement,which is an important performance of the intelligence and personalization in the intelligent education,particularly.At present,there are two types of mainstream test recommendation methods,including the question recommendation based on collaborative filtering and the question recommendation based on cognitive diagnosis.However,the former ignores the knowledge attribute of independent individuals,the latter is lack of the common evaluation.In order to improve the accuracy and efficiency of the question recommendation,comprehensive considering the knowledge attributes of the independent testing subjectand the knowledge commonality of the environment-like groups,this paper proposed a testing recommendation method based on collaborative filtering and cognitive diagnosis.Firstly,the proposed method designs a cognitive diagnosis model based on multi-level attributes scoring,which is used to model the subject’s answer.Then,the subject’s knowledge attribute is used for probabilistic matrix factorization to predict the potential answers.Finally,the appropriate questions are recommended to the subjects according to the information value.The testing recommendation comprehensively improves the interpretability and reliability that the experiment shows the method improves the efficacy by 20.35% and 2.5% respectively compared with collaborative filtering and cognitive diagnosis.
Traffic Congestion Prediction Based on Kernel Extreme Learning Machine Group Algorithm
XING Yi-ming, BAN Xiao-juan, LIU Xu, YIN Hang, SHEN Qing
Computer Science. 2019, 46 (11): 241-246.  doi:10.11896/jsjkx.191100507C
Abstract PDF(1804KB) ( 83 )   
References | Related Articles | Metrics
Prediction of urban traffic congestion is one of the important research contents of intelligent transportation system (ITS).At present,a lot of neural networks are introduced into the field of traffic forecasting and are widely used.However,the traditional neural network training is time-consuming,easy to fall into local optimal and over fitting.It has seriously hindered the large-scale application of neural network in the field of traffic forecasting.ELM is a new kind of single hidden layer feed-forward neural network,which has the advantages of fast training spead,stronggenera-lization ability and unique optimal solution.In this paper,the new algorithm named KELM-Group was proposed,which is composed of multiple KELM sub-models.KELM-Group algorithm enables each class of samples to achieve the global optimum,and the overall prediction accuracy can be higher than that of ELM.The experimental results show that the KELM-Group algorithm is faster than other popular machine learning algorithms.The accuracy rate of KELM-Group algorithm is 8% higher than that of the ELM.The results predicted by the KELM-Group algorithm are more consistent with the actual situation,and have great practical value.
Bionic Optimized Clustering Data Mining Algorithm Based on Cloud Computing Platform
SHEN Yan-ping, GU Su-hang, ZHENG Li-xia
Computer Science. 2019, 46 (11): 247-250.  doi:10.11896/jsjkx.190800042
Abstract PDF(1296KB) ( 81 )   
References | Related Articles | Metrics
In order to improve the validity of cloud computing platform data mining and the performance of data clustering,this paper combined bionic optimization algorithm with similar clustering to achieve cloud computing platform data clustering.In the process of solving the optimization function of similar clustering,wolf swarm optimization algorithm is used to locate the head wolf position to determine the cluster centers,so as to optimize and update the category centers.PBM and DB clustering effect evaluation methods were used to test the clustering effect,and wolf swarm optimization and similar clustering calculation were carried out continuously until the requirements of clustering index are met.Experiments results show that,compared with general clustering algorithms,wolf swarm optimization clustering algorithm has better clustering effect and faster convergence speed for cloud computing platform with large data volume and high data dimension.
Graphics ,Image & Pattern Recognition
Time Series Motif Discovery Algorithm of Variable Length Based on Domain Preference
WANG Yi-bo, PENG Guang-ju, HE Yuan-duo, WANG Ya-sha, ZHAO Jun-feng, WANG Jiang-tao
Computer Science. 2019, 46 (11): 251-259.  doi:10.11896/jsjkx.191100505C
Abstract PDF(2389KB) ( 116 )   
References | Related Articles | Metrics
With the development of ubiquitous computing,more and more sensors are installed in our daily applications.As a result,the demand for time series data processing is very high.The similar pattern which appears in time series data several times are called time series motif.Motif contains huge amounts of information in time series data.Motif discovery is one of the most important work in motif analysis.State-of-art motif discovery algorithm cannot find proper motif based on domain knowledge.As a result,such algorithm cannot find most valuable motif.Aiming at this problem,this paper used domain distance to evaluate the similarities of subsequences based on domain knowledge.By using the new distance,this paper developed a branching method to discovery motif with variable length.Several data from real life are used to test the performance of the algorithm.The results show that the proposed algorithm can find motif with domain knowledge accurately.
Image Denoising Algorithm Based on Fast and Adaptive Bidimensional Empirical Mode Decomposition
LIU Pei, JIA Jian, CHEN Li, AN Ying
Computer Science. 2019, 46 (11): 260-266.  doi:10.11896/jsjkx.190400159
Abstract PDF(4225KB) ( 169 )   
References | Related Articles | Metrics
In order to adaptively decompose the image and accurately describe the distribution state of the decomposition coefficients,a new image denoising algorithm based on fast and adaptive bidimensional empirical mode decomposition algorithm was proposed.Firstly,the algorithm performs fast and adaptive bidimensional empirical mode decomposition on the image.By determining the number of noise-dominated subband after decomposition,the noise-dominated subband coefficient distribution is further modeled by the normal inverse Gaussian model.Then the Bayesian maximum posteriori probability estimation theory is used to derive the corresponding threshold from the model.Finally,the optimal linear interpolation threshold function algorithm is used to complete the denoising.The simulation results show that for adding Gaussian white noise images of different standard deviation,the average signal-to-noise ratio is improved by 4.36dB,0.85dB,0.78dB and 0.48dB,respectively,compared with sym4 wavelet denoising,bivariate threshold denoising,pro-ximity algorithms for total variation,and overlapping group sparse total variation algorithm.Structural similarity index is also improved with different degrees,which shows it can effectively preserve more image details.The experimental results show that the proposed algorithm is superior to the comparison algorithms in terms of visual performance and evaluation index.
Person Re-identification Algorithm Based on Bidirectional KNN Ranking Optimization
BAO Zong-ming, GONG Sheng-rong, ZHONG Shan, YAN Ran, DAI Xing-hua
Computer Science. 2019, 46 (11): 267-271.  doi:10.11896/jsjkx.181001861
Abstract PDF(1462KB) ( 158 )   
References | Related Articles | Metrics
The imaging factors such as illumination,view,obstruction and noise would bring great changes to pedes-trian’s appearance under the cross-view condition in person re-identification,then it becomes very difficult to identify the target from candidates.Using the re-ranking algorithm can optimize the re-identification’s result,but it can make the task time-consuming and expensive.What’s more,it is easy to introduce the noise during the process of re-ranking,which in turn affects the accuracy of re-identification.To solve the problem,this paper presented a re-ranking method based on bidirectional KNN for person re-identification.First,it utilized the pre-training and fine-tuning strategy to extract the deep features of pedestrian.Then,it choosed an appropriate metric function (XQDA,KISSME) to measure the distance of features.Finally,accor-ding to the bidirectional KNN relation between the query and candidates,the Jaccard distance was calculated and aggregated with the original distance to guide the re-ranking.Experiments on the datasets of CUHK03,Market1501 and PRW show that the re-ranking algorithm proposed in this paper can improve the accuracy of re-identification on the basis of the original method,and the improvements are 12.2% and 13.4% in the two evaluation indexes of Rank1 and mAP respectively.The experimental data indicates that the re-identification algorithm based on bidirectional KNN can effectively reduce the probability of noise during the re-ranking,and then improve the accuracy of re-identification.
SSD Network Compression Fusing Weight and Filter Pruning
HAN Jia-lin, WANG Qi-qi, YANG Guo-wei, CHEN Jun, WANG Yi-zhong
Computer Science. 2019, 46 (11): 272-276.  doi:10.11896/jsjkx.180901630
Abstract PDF(1726KB) ( 93 )   
References | Related Articles | Metrics
Object detection is an important research direction in the field of computer vision.In recent years,deep lear-ning has achieved great breakthroughs in object detection which is based on the video.Deep learning has powerful ability of feature learning and feature representation.The ability enables it to automatically learn,extract and utilize relevant features.However,complex network structure makes the deep learning model have a large scale of parameter.The deep neural network is both computationally intensive and memory intensive.Single Shot MultiBox Detector300 (SSD300),a single-shot detector,produces markedly superior detection accuracy and speed by using a single deep neural network.But it is difficult to deploy it on object detection systems with limited hardware resources.To address this limitation,the fusing method of weight pruning and filter pruning was proposed to reduce the storage requirement and inference time required by neural networks without affecting its accuracy.Firstly,in order to reduce the number of excessive weight parameters in the model of deep neural network,the weight pruning method is proposed.Network connections is pruned,in which weight is unimportant.Then,to reduce the large computation in convolution layer,the redundant filters are pruned according to the percentage of effective weights in each layer.Finally,the pruned neural network is trained to restore its detection accuracy.To verify the effectiveness of the method,the SSD300 was validated on caffe which is the convolutional neural network framework.After compression and acceleration,the storage of SSD300 neural network required is 12.5MB and the detection speed is 50FPS.The fusion of weight and filter pruning achieves the result by 2× speed-up,which reduces the storage required by SSD300 by 8.4×,as little increase of error as possible.The fusing method of weight and filter pruning makes it possible for SSD300 to be embedded in intelligent systems to detect and track objects.
Method of Automatically Extracting Urban Water Bodies from High-resolution Images with Complex Background
WANG Wei-hong, CHEN Xiao, WU Wei, GAO Xing-yu
Computer Science. 2019, 46 (11): 277-283.  doi:10.11896/jsjkx.181001985
Abstract PDF(3109KB) ( 102 )   
References | Related Articles | Metrics
The distribution of urban water bodies is of great significance for people to understand the geographical phenomena such as the urban water circulation and the Heat-island Effect.It is common to obtain information by using high-resolution images for water extraction and water mapping.However,automatically extraction of water bodies by using the high-resolution images still is difficult for the complex background of the urban area,fewer spectral channels provided by the high-resolution images and the uneven distribution of water bodies in the images.This paper proposed an automatic extraction method of urban water bodies in complex background based on high-resolution images.First,adaptive threshold is selected for segmentation to gain the initial region of water,since water has a low gray value of the near infrared channel.Next,on the initial region,a buffering algorithm are used to obtain the target region of water extraction,and gauss mixture model and an expectation maximization algorithm is used to improve the distributionpara-meters of water.Then,the water bodies are extracted automatically using the maximum likelihood method with these parameters.As for the large number of shadow elements mixed in the rough extraction,a fusion features method is proposed to eliminate those noise points and obtain more accurate extraction result.The experiment results of water extraction in Jinshan show that the proposed method can effectively extract the structure of water bodies with small proportion in the experimental images,and perform well with high accuracy comparing to the commonly used automatic extraction algorithms.
Multi-modal Emotion Recognition Approach Based on Multi-task Learning
WU Liang-qing, ZHANG Dong, LI Shou-shan, CHEN Ying
Computer Science. 2019, 46 (11): 284-290.  doi:10.11896/jsjkx.180901665
Abstract PDF(2194KB) ( 382 )   
References | Related Articles | Metrics
Emotion analysis is a fundamental task of natural language processing(NLP),and the research on single modality (text modality) has been rather mature.However,for multi-modal contents such as videos which consist of three modalities including text,visual and acoustic modalities,additional modal information makes emotion analysis more challenging.In order to improve the performance of emotion recognition on multi-modal emotion datasets,this paper proposed a neural network approach based on multi-task learning.This approach simultaneously considers both intra-modality and inter-modality dynamics among three modalities.Specifically,three kinds of modality information are first preprocessed to extract the corresponding features.Secondly,private bidirectional LSTMs are constructed for each modality to acquire the intra-modality dynamics.Then,shared bidirectional LSTMs are built for modeling inter-modality dynamics,including bi-modal (text-visual,text-acoustic and visual-acoustic) and tri-modal interactions.Finally,the intra-modality dynamics and inter-modality dynamics obtained in the network are fused to get the final emotion recognition results through fully-connected layers and the Sigmoid layer.In the experiment of uni-modal emotion recognition,the proposed approach outperforms the state-of-the-art by 6.25%,0.75% and 2.38% in terms of text,visual and acoustic on average respectively.In addition,this approach can achieve average 65.67% in accuracy in multi-modal emotion recognition tasks,showing significant improvement compared with other baselines.
Fire Images Features Extraction Based on Improved Two-stream Convolution Network
XU Deng, HUANG Xiao-dong
Computer Science. 2019, 46 (11): 291-296.  doi:10.11896/jsjkx.180901640
Abstract PDF(2147KB) ( 271 )   
References | Related Articles | Metrics
Fire detection based on image processing technology is an important branch in the field of fire monitoring in recent years.Aiming at the fire detection of open environment,using the dynamic and static characteristics of smoke and flame generated during the fire,the two-stream convolutional neural network is used as the theoretical basis to detect the fire.The two-stream convolutional neural network uses spatial and temporal streams to extract spatial information and temporal information in the video respectively.However,the information in the early stage of the fire is weak and the features are not obvious enough.In order to improve the initial recognition rate,a spatial enhancement network was proposed as the spatial stream of the two-stream convolutional neural network to extract and enhance the spatial information of the video.The spatial enhancement network simultaneously convolves the current frame Vt and the previous frame Vt-1,subtracting the convolution features of the Vt image with the convolution features of the Vt-1 image,to preserve the difference of the convolution features,and adding the convolution features difference to the convolution features of the current frame Vt,thereby enhance the spatial features convolution of the current frame Vt-1.Temporal stream of two-stream convolutional network convolves the optical flow image Vt of the current frame to get the temporal features.Finally,the enhanced spatial and temporal features are fused to classify.The experimental results show that the improved two-stream convolutional network has a 6.2% higher recognition rate than the original two-stream convolutional network,and achieved 92.15% recognition rate on the public dataset,indicating the effectiveness and superiority of the proposed method.Comparing with other methods,the network structure is designed lower but achieves good results,improves the identification accuracy of fire and smoke as well as realizes the early warning of fire,shorten detection time.
Interdiscipline & Frontier
WiCount:A Crowd Counting Method Based on WiFi Channel State Information
DING Ya-san, GUO Bin, XIN Tong, WANG Pei, WANG Zhu, YU Zhi-wen
Computer Science. 2019, 46 (11): 297-303.  doi:10.11896/jsjkx.191100506C
Abstract PDF(2584KB) ( 293 )   
References | Related Articles | Metrics
Crowd counting is the process of monitoring the number of people in a certain area,which is crucial in traffic supervision,etc.For example,counting people waiting in lines at airports or retail stores could be used for improving the service.At present,some methods based on videos (or images) and wearable devices have been proposed,but there are some shortcomings in these schemes.For example,the camera can only monitor within the range of sight distance,and wearable devices need people to wear them consciously.Some scholars have made use of radar related technology torealize the number,but its cost is very high.In this paper,an indoor crowd counting scheme,WiCount,based on WiFi signals was proposed.WiCount aims at a fine-grained indoor people counting scheme,which can accurately identify the number of people at different positions.According to the relationship between the number of indoor people and the amplitudes fluctuation of CSI,features are extracted,which are contributed to mitigate the difference of CSI data produced by the same number of people in distinct positions,and then three classifiers (SVM,KNN,BP Neural Network) are trained to identify the number of people in the monitoring area.Prototype systems is implemented in a laboratory and a meeting room respectively,and the recognition is fine when the number of people is on the small side.In the laboratory,the accuracy is up to 90% in the case of no more than 4 persons.In the meeting room,the results show that no matter where people move,the accuracy can reach 89.58% in the case of no more than 2 persons.
Training-free Human Respiration Sensing Based on Wi-Fi Signal
YU Yi-ran, CHANG Jun, WU Liu-fan, ZHANG Yong-hong
Computer Science. 2019, 46 (11): 304-308.  doi:10.11896/jsjkx.190600143
Abstract PDF(1962KB) ( 93 )   
References | Related Articles | Metrics
With the rapid development of wireless communication technology,Wi-Fi has been widely used in public and private fields.Non-invasive breath detection technology based on wireless technology has a broad application prospect in the field of smart home.Considering that the existing solutions are difficult to explain the huge performance differences in different scenarios,this paper introduced the Fresnel edge diffraction model in the free space and designd a training free breathing detection sensing based on Wi-Fi signals.Firstly,we introduced the Fresnel Zone knife-edge diffraction model in free space,then verified the diffraction propagation characteristics of Wi-Fi signals in indoor environment.Se-condly,we accurately quantified the relationship between diffraction gain and micro thoracic displacement in human respiration,which Not only explains why Wi-Fi devices can be used to detect human breathing,but also demonstrates where is easier to detect.Finally,respiratory rate is estimated from RSS by fast Fourier transform (FFT).The algorithm in this paper can clearly know the distribution of good and bad positions of breath detection,and for good positions,the accuracy of breath estimation can reach 93.8%.Experiment results show that using a pair of transceivers makes centimeter-scale breathing perception possible and it is expected to provide a ubiquitous respiratory detection solution through a popular Wi-Fi infrastructure.
Cloud Computing Resource Scheduling Strategy Based on Improved Bacterial Foraging Algorithm
ZHAO Hong-wei, TIAN Li-wei
Computer Science. 2019, 46 (11): 309-314.  doi:10.11896/jsjkx.181002000
Abstract PDF(1996KB) ( 124 )   
References | Related Articles | Metrics
As one of the core problems of cloud computing,the efficiency of scheduling algorithm has a direct impact on the operation capacity of the system.Swarm intelligence algorithm,with good coordination and overall stability,is one kind of swarm intelligence algorithms which imitates swarm intelligence in the process of evolution swarm.This paper presented a calculation method of bacteria foraging algorithm applied to cloud computing resource scheduling algorithm,which can be used to control the node allocation of cloud computing resource scheduling by using bacterial swarm algorithm to copy and perish the nodes.According to the problem of too much resource change interval caused by the random selection chemotax in the traditional flora swarm algorithm,the bacteria foraging CBFO optimization algorithm based on Quorum Sensing mechanism and the MPSOBS optimization algorithm introducing bacteria chemotaxis action in the process of group collaboration were proposed in this paper.According to the environment around the nodes and the situation of the whole flora,the chemotaxis factor is selected to make the process of chemotaxis more accurate,which is implemented on the cloud computing platform.The simulation results show that the proposed algorithm is more efficient than the BFO algorithm in terms of task execution time,system load balancing and resource service quality,and can improve the service quality of cloud applications while improving resource utilization.
Load Balancing Scheduling Optimization of Cloud Workflow Using Improved Shuffled Frog Leaping Algorithm
XU Jun, XIANG Qian-hong, XIAO Gang
Computer Science. 2019, 46 (11): 315-322.  doi:10.11896/jsjkx.181001866
Abstract PDF(1834KB) ( 132 )   
References | Related Articles | Metrics
In instance-intensive and open cloud environments,workflow scheduling always suffers from frequent calls of the cheap and high-quality resources,resulting in poor scheduling efficiency and disruption of stability.In addition,unlike general task scheduling,workflow tasks usually have associated dependencies,which greatly increase the complexity of task assignment.Aiming at the imbalance of load between cloud virtual machines,a workflow hierarchical scheduling model was proposed,which is hierarchically divided according to task priorities so as to alleviate virtual machine load pressure.Besides,to optimize the shuffled frog leaping algorithm(ISFLA),the time greedy strategy is applied to initia-lize population,as a result,improving the search efficiency.Then,by enhancing the position of best solutions locally,a reconstruction strategy was put forward to go out of the dilemma of local optimum.Finally,the experimental results in cloud workflow scheduling show that the improved shuffled frog leaping algorithm can optimize load balance degree and is more effective in task processing as well as searching compared with the traditional shuffled frog leaping algorithm and particle swarm optimization.
Bearing Fault Diagnosis Method Based on Variational Bayes
WANG Yan, LUO Qian, DENG Hui
Computer Science. 2019, 46 (11): 323-327.  doi:10.11896/jsjkx.180901719
Abstract PDF(1442KB) ( 91 )   
References | Related Articles | Metrics
Rolling bearings are common parts in rotating mechanical structures and can cause significant damage if they fail.With the advent of the era of big data,modern intelligent diagnostic methods are widely used in bearing fault diagnosis.Aiming at the problems existing in the intelligent diagnosis method,this paper introduced the statistical model into the bearing fault diagnosis,and proposed a fault diagnosis method based on the variational Bayesian.The method performs local feature scale decomposition on the bearing vibration signal to obtain several intrinsic scale components and extracts the time domain feature composition feature set.The feature set training is used to generate the mixed multidimensional Gaussian distribution model based on variational Bayes,and the different bearings are calculated.The probability of failure is to achieve fault diagnosis.The experimental results show that the diagnostic accuracy rate is 99.6%.Compared with the bearing diagnosis method based on support vector machine,diagnostic accuracy rate is up to 39.6%.The proposed method can comprehensively and effectively diagnose rolling bearing faults,and has a good diagnostic effect on high-dimensional complex fault data.
Study on Optimization of Quantization Algorithm in ADMM Decoding Algorithm Based on Lookup Table
LIU Hua-jun, TANG Shi-di, ZHANG Di-ke, XIA Qiao-qiao
Computer Science. 2019, 46 (11): 328-333.  doi:10.11896/jsjkx.181001871
Abstract PDF(2377KB) ( 60 )   
References | Related Articles | Metrics
In the linear programming decoding based on ADMM,the Euclidean projection calculation of the projection vector to the check multi-cellular body is the most complex and time-consuming part.The ADMM-LDPC decoding algorithm based on look-up tables replace the time complexity of algorithm with simple table look-up operations to simplify the projection process and improve the efficiency of the algorithm,however it expends much memory consumption.La-ter,the researchers proposed the nonuniform quantization method,which can decrease the memory consumption effectively,but the computation complexity of the quantitative method is too high,so that the way is difficult to achieve when the number of segments is too much.For this problem,this paper proposed a new nonuniform quantization method.Firstly,for different code-words,the distribution characteristics of elements in the vector are calculated to be projected under different SNRs conditions by experiment,its distribution rules are explored and the corresponding function is designed as the quantitative mapping relation.Then,differential evolution algorithm is used to optimize the parameters of the function,the optimal quantization schema under the function is obtained,and the quantization function is finally determined.The simulation results show that,compared with the existing quantitative methods,the nonuniform method proposed in this paper has the advantage of not being affected by the number of quantization segment,precision and otherfactors.What’s more,it achieves about 0.05dB performance improvement for different code words in high SNRs.
Approach for Mining Block Structure Process from Complex Log Using Log Partitioning
DUAN Rui, FANG Huan, ZHAN Yue
Computer Science. 2019, 46 (11): 334-339.  doi:10.11896/jsjkx.180901710
Abstract PDF(1544KB) ( 74 )   
References | Related Articles | Metrics
With the development of enterprises,more and more logs are generated and recorded by the system,and the process of mining block structures from cumbersome and complicated logs becomes more challenging.This paper proposed an approach of vertically dividing logs,greatly reducing the number of instances of each log partition,and shorte-ning the length of each trace to process complex logs and mining accurate models from them.The basis of log division is activity division.Firstly,based on the idea of behavioral association,the concept of common transition is proposed torea-lize the aggregation division of interrelated activities.Then,from the perspective of the number of common transitions in the log,the activity set is divided by mutually different and interleaved methods,thereby realizing the division of mo-dules and logs.The proposed module and log partitioning method can be iterated until that the log partitioning is simple enough.Finally,a block structure is mined from each divided simple log,and a reasonable overall system model is formed by combining block structures.The feasibility of the proposed method is verified by Prom experiment.