Computer Science

Review for Deep Learning Based on Medical Imaging Diagnosis

ZHANG Qiao-li, ZHAO Di and CHI Xue-bin

Computer Science. 2017, 44 (Z11): 1-7. doi:10.11896/j.issn.1002-137X.2017.11A.001

Abstract

PDF(1567KB) ( 3246 )

References | Related Articles | Metrics

At present,the various modalities of medical image data accumulate rapidly,bringing great challenges to doctors who diagnose disease through traditional medical image analysis methods.Deep learning method has gained great success and become more and more popular in the computer vision field.All that case provides new chances for automaticmedical image analysis and makes high precisely computer-aided disease diagnosis possible.In this paper,we reviewed state-of-the-art research progress of deep learning in the medical image field.Firstly,the method of deep learning and its application in the field of medical imaging are introduced.Then attention is focused on specific research progress of deep learning method in several typical and popular disease.Finally,the tendency of this research field is summarized,and then the existing problems and recommendations are put forward.

Text Extraction in Video and Images:A Review

JIANG Meng-di, CHENG Jiang-hua, CHEN Ming-hui and KU Xi-shu

Computer Science. 2017, 44 (Z11): 8-18. doi:10.11896/j.issn.1002-137X.2017.11A.002

Abstract

PDF(2184KB) ( 2179 )

References | Related Articles | Metrics

Text extraction in video and images has important application value.Big data era brought urgent demands of huge amounts of information retrieval,many text extraction methods have been proposed in recent years.In this paper,we reviewed text extraction methods from video and images.First,we classified the course of text extraction into two steps:text region detection and localization,text segmentation.Then,some text region detection and localization and text segmentation algorithms have been discussed regarding their application fields and their advantages and disadvantages.Finally,we discussed benchmark data and performance evaluation,and pointed out the promising directions for future research.

Review of Implicit Surface Reconstruction from Point Cloud Dataset

XU Li-min and WU Gang

Computer Science. 2017, 44 (Z11): 19-23. doi:10.11896/j.issn.1002-137X.2017.11A.003

Abstract

PDF(1317KB) ( 2292 )

References | Related Articles | Metrics

The surface reconstruction of point cloud data is to reconstruct the surface of 3D objects on the scattered data points of the scanning equipment.It is widely used in computer animation,target recognition,data visualization and geographic information system.The surface reconstruction based on implicit function is an important method of surface reconstruction of point cloud datasets due to its ability to reconstruct holes and cracks,and then do not need splicing and smoothing.This paper summarized some main implicit surface reconstruction methods,analyzed and compared the advantages and disadvantages of implicit model and corresponding surface reconstruction algorithm.Finally,the existing problems and the future development direction of implicit surface reconstruction were analyzed and discussed.

Survey on Monitoring Techniques for Data Abnormalities

WU Jing-feng, JIN Wei-dong and TANG Peng

Computer Science. 2017, 44 (Z11): 24-28. doi:10.11896/j.issn.1002-137X.2017.11A.004

Abstract

PDF(1280KB) ( 3662 )

References | Related Articles | Metrics

In the current environment of large data,anomaly data is more difficult to obtain than normal data,and is more important.The purpose of anomaly detection is to detect activity data from normal subjects.Anomaly detection is widely applied in many fields,such as machine fault detection,data mining and disease detection and intrusion detection.Based on a large number of anomaly detection methods at present,this paper mainly discussed the existence of anomaly data,classified the major anomaly detection methods according to this framework,and put forward the advantages and disadvantages of these methods.Finally,we focused on the large data anomaly detection methods based on the deep learning,and introduced different methods and related applications and future research hotpots respectively.

Application Survey of Artificial Intelligence in Neurology

LI Shi-yu, WANG Feng, CAO Bin and MEI Qi

Computer Science. 2017, 44 (Z11): 29-32. doi:10.11896/j.issn.1002-137X.2017.11A.005

Abstract

PDF(1262KB) ( 2021 )

References | Related Articles | Metrics

Artificial intelligence affects all aspects of people’s lives,and medical treatment has become one of the most popular areas of artificial intelligence.More and more artificial intelligence equipments are used to assist doctors in the diagnosis and treatment.The application of artificial intelligence in neurology was reviewed,especially for the diagnosis of Parkinson’s disease and Alzheimer’s disease.Firstly,the development history,classification and application status of artificial intelligence were expounded.Secondly,the research status of artificial intelligence diagnosis of Parkinson’s di-sease and Alzheimer’s disease was summarized,and the key technologies used to diagnosis Parkinson’s disease and Alzheimer’s disease were analyzed.Finally,The technology in the application of medicine was summarized,and the importance of the application of artificial intelligence in the medical field was clarified.The future research direction of artificial intelligence in the application of neurology was prospected.

Survey on Reliability Estimation Methods of Sequential Circuit in Height-level

OUYANG Cheng-tian, CHEN Li-li and WANG Xi

Computer Science. 2017, 44 (Z11): 33-38. doi:10.11896/j.issn.1002-137X.2017.11A.006

Abstract

PDF(1471KB) ( 1334 )

References | Related Articles | Metrics

Reliability of sequential circuits is emerging as an important concern in scaled electronic technologies.In this paper,a survey of the research progress of the high-level reliability analysis for sequential circuits was given.Specially,we focused on Bayesian reliability analysis,multiple-pass reliability analysis and reliability estimation of sequential circuit based on probabilistic transfer matrix.And these analysis methods of sequential circuit were selected for experiment on the ISCAS 89 benchmark circuits.Research results and experimental results show that the abstraction level of circuit is higher,the accuracy of the results is lower,and the time overhead will be less.In the same abstraction level,the simulation methods have high accuracy,but also have more runtime,and analytical methods have low time overhead,but less accurate.

Repetitive Pattern Recognition Algorithms and Applications in Web Information Extraction and Clustering Analysis

Munina YUSUFU and Gulina YUSUFU

Computer Science. 2017, 44 (Z11): 39-45. doi:10.11896/j.issn.1002-137X.2017.11A.007

Abstract

PDF(1415KB) ( 1466 )

References | Related Articles | Metrics

Detection of repetitive patterns in sequences and their applications have become fundamental research areas in data mining.It is a very important means for extracting useful information from sequences.In recent years,many works have been conducted focusing on the definitions of repetitive patterns,efficient recognition algorithm designs,and applications in the relevant areas.In this paper,the classification and characteristics of repetitive patterns in sequence were briefly described,and the commonly used data structures in the algorithms were discussed.Recent studies on the applications of detection algorithms in relevant fields and their main design ideas were reviewed by discussing and evaluating certain aspects.These aspects include the field knowledge and constraints,identifying results,scalabilities of algorithms and existing main problems.Application areas include Web information extraction,Web document feature extraction and clustering algorithms,and related Uyghur language information processing.Finally,some challenges on detection algorithms of repetitive patterns in sequences and applications in various fields were discussed and future research trends were also explored.

Ordering Recommender Algorithm Based on Consumers’ Behavior

DING Dang, ZHANG Zhi-fei, MIAO Duo-qian and CHEN Yue-feng

Computer Science. 2017, 44 (Z11): 46-50. doi:10.11896/j.issn.1002-137X.2017.11A.008

Abstract

PDF(1390KB) ( 1858 )

References | Related Articles | Metrics

With the development of e-commerce,most of the existing catering management system lags consumers and managers’ need.An effective approach is to apply recommendation systems to catering management,and to provide ordering recommendations according to consumers’ behavior data.As for cold start problems that may arise in the recommending process,the ordering recommender system based on consumers’ behavior was proposed,containing three re-commendation engines,which are frequency statistics,association rules and Markov chain .Experiments on ordering data of real restaurants achieve a satisfactory result,and get a weight combination of three recommendation engine:(0.2167,0.5167,0.2666),and the best recommending length under that weight:3.

Application of BP Neural Network Based on Newly Improved Particle Swarm Optimization Algorithm in Fitting Nonlinear Function

LIN Yu-feng, DENG Hong-min and SHI Xing-yu

Computer Science. 2017, 44 (Z11): 51-54. doi:10.11896/j.issn.1002-137X.2017.11A.009

Abstract

PDF(1297KB) ( 1613 )

References | Related Articles | Metrics

A newly improved particle swarm optimization algorithm for the optimization of BP neural network was introduced to solve the problem of large error in fitting the nonlinear function.In this algorithm,a new model based on the newly improved particle swarm optimization algorithm is established through respectively changing the weight non-li-nearly and learning factor linearly,then it is applied to non-linear function fitting by combining with BP neural network.The results show that the newly improved particle swarm optimization algorithm can more rationally and effectively boost the fitting ability of BP neural network,and improve the accuracy of the fitting.

Research on Fuzzy Matching Duplicate Checking Algorithm Based on Matrix Model of Word Segmentation

LI Cheng-long, YANG Dong-ju and HAN Yan-bo

Computer Science. 2017, 44 (Z11): 55-60. doi:10.11896/j.issn.1002-137X.2017.11A.010

Abstract

PDF(1502KB) ( 3208 )

References | Related Articles | Metrics

Aiming at the need of Chinese text duplicate checking,based on the result of word segmentation,we converted target text and sample text into matrix model of word segmentation,then scanned and analyzed matrix to get the result.Therefore an algorithm of duplicate checking was developed,and the usefulness of the method was demonstrated by practical examples.

Analysis and Prediction on Rebar Price Based on Multiple Linear Regression Model

CHEN Hai-peng, LU Xu-wang, SHEN Xuan-jing and YANG Ying-zhuo

Computer Science. 2017, 44 (Z11): 61-64. doi:10.11896/j.issn.1002-137X.2017.11A.011

Abstract

PDF(1380KB) ( 2349 )

References | Related Articles | Metrics

A kind of rebar price analysis as well as prediction model based on multiple linear regression analysis was proposed by means of analyzing the upstream and downstream relationship of rebar industrial chain in futures black line variety.Firstly,the data of major factors influencing rebar price is collected,including coke futures settlement price,coking coal futures settlement price,iron ore futures settlement price,hot rolled futures settlement price,and central parity rate of RMB to USD.Later,these influencing factors are analyzed through scatter diagram and rend line to determine influencing factors.The multiple linear regression model based on least square method is constructed by virtue of SPSS and NCSS,and the collected data.Meanwhile,the collinearity among independent variables are moved through ridge regression to obtain revised model.At last,this model is applied to carry out accurate prediction of rebar price on trade day in the next month.The experiment indicates that the fitting degree of this model is higher with certain practicability.

New Intelligent Prediction of Chronic Liver Disease Based on Principal Component Machine Learning Algorithm

CHANG Bing-guo, LI Yu-qin, FENG Zhi-chao and YAO Shan-hu

Computer Science. 2017, 44 (Z11): 65-67. doi:10.11896/j.issn.1002-137X.2017.11A.012

Abstract

PDF(1302KB) ( 1400 )

References | Related Articles | Metrics

Using new information technology to predict the mechanism and characteristics of chronic liver disease is an effective way to improve its diagnosis.In this paper,we used the principal component analysis (PCA) of the machine learning algorithm to reduce the dimensional indicators of chronic liver disease,combined with neural network learning to build a new intelligent prediction of chronic liver disease (IPCLD).The experiment studied 125 data sets of 20-dimensional indicators of chronic liver disease,used receiver operating characteristic (ROC) curve to select 13-dimensional more sensitive indicators,further reduces the dimension down to 5 by PCA.The neural network is trained with 115 data sets,and the remaining 10 data sets are used as test data sets.Compared with being trained by original data,the IPCLD improves 15.07% prediction accuracy and reduces the complexity.

Study on Abnormal Diagnosis of Moving ECG Signals Based on Unsupervised Learning

LI Feng and XIE Si-hong

Computer Science. 2017, 44 (Z11): 68-71. doi:10.11896/j.issn.1002-137X.2017.11A.013

Abstract

PDF(1413KB) ( 1714 )

References | Related Articles | Metrics

To diagnosis the abnormal mobile ECG signals,a method based on unsupervised learning was proposed.The ECG data is classified by hierarchical clustering,and by combining the priority diagnosis method of the feature quantity, the complexity of time and the complexity of space consuming are effectively reduced.At last,the analysis of the examples verifies the methods of this paper.

UAV Navigation Algorithm Research Based on EB-RRT*

CHEN Jin-yin, LI Yu-wei and DU Wen-yao

Computer Science. 2017, 44 (Z11): 72-79. doi:10.11896/j.issn.1002-137X.2017.11A.014

Abstract

PDF(1767KB) ( 1147 )

References | Related Articles | Metrics

With the wide application of UAV,automatic navigation capability of UAV becomes more important.UAV navigation algorithm was defined to plan a collision free and smooth path from start position to destination position in known maps.Aiming at three problems of current navigation algorithms including low convergence rate,long searching time and navigation path couldn’t be applied to real UAV,EB-RRT* (Efficient B-RRT*) algorithm was proposed.A self-adaptive obstacle avoidance strategy was designed to speed up convergence rate of traditional navigation algorithm by reducing memory cost.Grid based segmentation mechanism was put forward to reduce searching time for path planning.Suitable down sampling and three times Bessel interpolation formula were adopted to smooth final path for practical UAV applications.Several simulation maps were used to testify the performances of proposed algorithms compared with other classic algorithms.

Algorithm of Importance Ranking for Influencing Factors of Website Service Quality Based on PageRank

QI Yu-dong, HE Cheng and YUAN Wei

Computer Science. 2017, 44 (Z11): 80-83. doi:10.11896/j.issn.1002-137X.2017.11A.015

Abstract

PDF(1186KB) ( 1100 )

References | Related Articles | Metrics

In this paper,factors were filtered through the Delphi method.But in the analysis of the relationship between the various factors,PageRank voting ideas look the interaction of the factors which affect the quality of web service as a link with each webpage,and this method uses this interaction as votes and calculates the affected value by internal relationship of each factors.The sequence of factors which affects the quality of web service by importance will be got eventually,and this method offers an example and reference for evaluation of quality of the web service.

Research of Path Planning in Multi-AGV System

TAI Ying-peng, XING Ke-xin, LIN Ye-gui and ZHANG Wen-an

Computer Science. 2017, 44 (Z11): 84-87. doi:10.11896/j.issn.1002-137X.2017.11A.016

Abstract

PDF(1269KB) ( 4802 )

References | Related Articles | Metrics

In this paper,a dynamic routing method based on time window model was presented for dealing with multi-AGV path planning problem in warehouse.Firstly,A* algorithm is used for completing path routing of multiple AGV.Secondly,the time for AGV passing through the path node’s is calculated.Multi-AGV conflict issues is solved by inserting the time window into path and updating time window.Furthermore,through dynamic allocation of priority for the multi-AGV,the efficiency of system is enhanced.Finally,when the obstacles are in the path,by dynamically changing the path weight,real-time obstacle avoidance is achieved.The simulation results show that the proposed algorithm can effectively avoid collision under the condition of optimal path,the method can not only improve the system efficiency,but also has good adaptability and robustness in dynamic environment.

Fast Incremental Learning Algorithm of SVM with Locality Sensitive Hashing

YAO Ming-hai, LIN Xuan-min and WANG Xian-bao

Computer Science. 2017, 44 (Z11): 88-91. doi:10.11896/j.issn.1002-137X.2017.11A.017

Abstract

PDF(1231KB) ( 1452 )

References | Related Articles | Metrics

In order to improve the training speed and the classification accuracy in large scale high dimension data,a new incremental learning algorithm of SVM with LSH was proposed.It uses the LSH algorithm,which can seek similar data fast in a large scale and high dimension data,to filter out the incremental samples which may become SVs on the basis of the SVM algorithm.Then it makes the selected samples and the existing SVs as a basis for the following training.We took advantages of the multiple data sets to validate the algorithm.Experiments show that this new algorithm can improve the speed of the incremental training learning in large scale data with the effective accuracy.

Research of Text Sentiment Classification Based on Improved Semantic Comprehension

WANG Ri-hong, CUI Xing-mei, ZHOU Wei, WANG Cheng-long and LI Yong-jun

Computer Science. 2017, 44 (Z11): 92-97. doi:10.11896/j.issn.1002-137X.2017.11A.018

Abstract

PDF(1429KB) ( 1576 )

References | Related Articles | Metrics

Text classification has a wide range of applications in information retrieval,Web automatic document classification,digital library,automatic abstracting,document organization and management.An improved text sentiment classification method was put forward based on semantic understanding.Emotional sememe was joined to revise the definition in the emotional similarity calculation and the development of emotional phrases’ sentiment was combined.Focusing on emotional words and negative words,the degree of adverbs combining form of analysis,and module complex negative word and adverb were put forward.Combining with the use of conjunctions as the standard of the sentence for emotional tendencies classified processing,a text propensity algorithm was given to judge the text sentiment and classify the text.Experimental results show that the classification results with the method improve when comparing with the previous algorithm.

Cost-sensitive Random Forest Classifier with New Impurity Measurement

SHI Yan-wen and WANG Hong-jie

Computer Science. 2017, 44 (Z11): 98-101. doi:10.11896/j.issn.1002-137X.2017.11A.019

Abstract

PDF(1364KB) ( 1449 )

References | Related Articles | Metrics

For the problem of effective classification on imbalanced data sets,a classifier combining cost-sensitive learning and random forest algorithm is proposed.Firstly,a new impurity measure is proposed,taking into account not only the total cost of the decision tree,but also the cost difference of the same node for different samples.Then,the random forest algorithm is executed,K times sampling for the data set is performed,and K basic classifiers are built.Then,the decision tree is constructed by the classification regression tree (CART) algorithm based on the proposed impurity measure,so as to form the decision tree forest.Finally,the random forest algorithm makes the data classification decision by voting mechanism.In the UCI database,compared with the traditional random forest and the existing cost-sensitive random forest