计算机科学 ›› 2020, Vol. 47 ›› Issue (6A): 494-499.doi: 10.11896/JsJkx.190900016

• 数据库 & 大数据 & 数据科学 • 上一篇    下一篇

基于牛顿法的自适应高阶评分距离推荐模型研究

邹海涛, 郑尚, 王琦, 于化龙, 高尚   

  1. 江苏科技大学计算机学院 江苏 镇江 212003
  • 发布日期:2020-07-07
  • 通讯作者: 高尚(gao_shang@sohu.com)
  • 作者简介:nkroben@outlook.com
  • 基金资助:
    江苏省高等学校自然科学研究基金(18JBK520011);镇江市重点研发计划-社会发展(SH2019021)

Adaptive High-order Rating Distance Recommendation Model Based on Newton Optimization

ZOU Hai-tao, ZHENG Shang, WANG Qi, YU Hua-long and GAO Shang   

  1. School of Computer Science,Jiangsu University of Science and Technology,ZhenJiang,Jiangsu 212003,China
  • Published:2020-07-07
  • About author:ZOU Hai-tao, born in 1984, Ph.D, lecturer.His main research interests include data mining and information retrieval.
    GAO Shang, born in 1972, Ph.D, professor, is a member of China Computer Federation.His main research interests include intelligent computing and pattern recognition.
  • Supported by:
    This work was suppored by the Natural Science Foundation of the Higher Education Institutions of Jiangsu Province, China (18JBK520011) and Primary Research & Developement Plan (Social development) of ZhenJiang City (SH2019021).

摘要: 现有的一些算法引入了隐语义模型克服数据稀缺带来的问题,为用户提供更有效的推荐。一般情况下,这些方法通过线性组合若干多项式,引入相应参数平衡各个部分比重,以构造优化函数,最终达到最小评分误差或实现最大的偏好等目的。经典模型通常只考虑用户对某一产品的预测评分和实际评分差异(即,一阶评分距离),忽略了其在不同产品上的预测评分与实际评分之间的差值 (即,二阶评分距离)。因此,高阶评分距离模型同时将两种距离集成到算法之中,并使用随机梯度下降法求解目标函数。可是,上述优化函数中的相关参数往往是手动设置,而且随机梯度下降法求解目标函数的收敛速度较慢,使得该模型缺乏灵活性,也增加了时间消耗。为了提高模型的适应性和效率,文中提出了一种融合归一化函数的自适应高阶评价距离模型,并利用牛顿法求解改进后的高阶评分距离凸优化函数。此方法不仅移除了若干静态参数,而且加快了优化函数的收敛速度。提出的模型具有坚实的理论支持,经过3个实际数据集的实验结果表明,此模型具有较好的预测精度和运行效率。

关键词: 推荐系统, 隐语义模型, 牛顿法, 凸优化, 高价评分距离

Abstract: Some existing recommendation algorithms introduce latent factor model to overcome the problems caused by data scarcity,so as to provide more effective recommendations for users.In general,those methods construct an optimization function to achieve the minimum rating error or maximum preference,etc,by integrating several polynomials with the corresponding parameters to balance each part,and use stochastic gradient descent to solve this function.Nevertheless,the above mentioned models only consider the difference between the estimated and real ratings of the same user-item pair (i.e.,the first-order rating distance),and ignore the difference between the estimated and real ratings of the same user across different items (i.e.,the second-order rating distance).Hence,high-order rating distance model,HoORaYs,with good accuracy in terms of item ranking and predictive ratings which takes these two kinds of distances into account is proposed.Unfortunately,this model still has some flaws in adap-tability and efficiency due to its manually setting parameters,its non-convergence.Aiming at improving the recommendation adap-tability and efficiency,an adaptive high-order rating distance model which integrates a data scale sensitive function is proposed.It utilizes Newton method to solve the convex optimization problem about rating distance.This method not only eliminates manually setting parameters,but also accelerates the optimization function convergence speed.The proposed model has a solid theoretical support.Experiments on three real datasets show that,it has good prediction accuracy and operation efficiency.

Key words: Recommender systems, Latent factor model, Newton method, Convex optimization, High-order rating distance

中图分类号: 

  • TP181
[1] KABBUR S,NING X,KARYPIS G.FISM:factored item similarity models for top-N recommender systems//Internatio-nal Conference on Knowledge Discovery & Data Mining,2013:659-667.
[2] ZHOU K,ZHA H.Learning binary codes for collaborative filtering//International Conference on Knowledge Discovery and Data Mining.2012:498-506.
[3] ZHANG Z,WANG Q,RUAN L,et al.Preference preserving hashing for efficient recommendation//International Confe-rence on Research and Development in Information Retrieval.2014:183-192.
[4] BHARGAVA P,PHAN T,ZHOU J,et al.Who,What,When,and Where:Multi-Dimensional Collaborative Recommendations Using Tensor Factorization on Sparse User-Generated Data//International Conference on World Wide Web.2015:130-140.
[5] CHENG Z,YING D,LEI Z,et al.Aspect-aware latent factor model:rating prediction with ratings and reviews//International Conference on World Wide Web.2018:639-648.
[6] XU J,YAO Y,TONG H,et al.HoORaYs:High-order optimization of rating distance for recommender systems//Procee-dings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD ’17).2017:525-534.
[7] ZOU H,GONG Z,HU W.Identifying diverse reviews about products.World Wide Web,2017,20(2):351-369.
[8] LIANG D,ALTOSAAR J,CHARLIN L,et al.Factorization meets the item embedding:regularizing matrix factorization with item co-occurrence//International Conference on Recommender Systems.2016:59-66.
[9] GEMULLA R,NIJKAMP E,HAAS P J,et al.Large-scale matrix factorization with distributed stochastic gradient descent//International Conference on Knowledge Discovery and Data Mining,2011:69-77.
[10] XU C,ZHENG Q,ZHANG Y,et al.Learning to rank features for recommendation over multiple categories//International Conference on Research and Development in Information Retrieval.2016:305-314.
[11] GUO G,ZHANG J,YORKE-SMITH N.Trustsvd:Collabora tive filtering with both the explicit and implicit influence of user trust and of item ratings//Twenty-ninth AAAI Conference on Artificial Intelligence.2015:123-129.
[12] OH J,HAN W S,YU H,et al.Fast and robust parallel SGD matrix factorization//International Conference onKnow-ledge Discovery and Data Mining.2015:865-874.
[13] WU L,HSIEH C J,SHARPNACK J.Large-scale collaborative ranking in near-linear time//International Conference on Knowledge Discovery and Data Mining.2017:515-524.
[14] HESTENES M R,STIEFEL E.Methods of conJugate gradients for solving linear systems.J.Res.Nat.Bur.Stand,1952,49(6):409-436.
[15] RODIN E Y.Nonlinear programming analysis and methods.Computers & Mathematics with Applications,1977,3(2):151-151.
[1] 李太松,贺泽宇,王冰,颜永红,唐向红. 基于循环时间卷积网络的序列流推荐算法[J]. 计算机科学, 2020, 47(3): 103-109.
[2] 冯晨娇,梁吉业,宋鹏,王智强. 基于极端评分行为的相似度计算[J]. 计算机科学, 2020, 47(2): 31-36.
[3] 余敦辉, 成涛, 袁旭. 基于排序学习的软件众包任务推荐算法[J]. 计算机科学, 2020, 47(12): 106-113.
[4] 黄超然. 基于显式反馈协同过滤算法的偏好与共性平衡[J]. 计算机科学, 2020, 47(11A): 471-473.
[5] 罗鹏宇, 吴乐, 吕扬, 袁堃平, 洪日昌. 基于时序推理的分层会话感知推荐模型[J]. 计算机科学, 2020, 47(11): 73-79.
[6] 张艳红, 张春光, 周湘贞, 王怡鸥. 项目多属性模糊联合的多样性视频推荐算法[J]. 计算机科学, 2019, 46(8): 78-83.
[7] 郭旭, 朱敬华. 基于用户向量化表示和注意力机制的深度神经网络推荐模型[J]. 计算机科学, 2019, 46(8): 111-115.
[8] 王雅慧, 刘博, 袁晓彤. 基于近似牛顿法的分布式卷积神经网络训练[J]. 计算机科学, 2019, 46(7): 180-185.
[9] 王旭, 庞巍, 王喆. 异构信息网络中基于元结构的协同过滤算法[J]. 计算机科学, 2019, 46(6A): 397-401.
[10] 刘晴晴, 罗永龙, 汪逸飞, 郑孝遥, 陈文. 基于SVD填充的混合推荐算法[J]. 计算机科学, 2019, 46(6A): 468-472.
[11] 石晓玲, 陈芷, 杨立功, 沈伟. 基于自适应样本权重的矩阵分解推荐算法[J]. 计算机科学, 2019, 46(6A): 488-492.
[12] 陈俊航, 徐小平, 杨恒泓. 基于Seq2seq模型的推荐应用研究[J]. 计算机科学, 2019, 46(6A): 493-496.
[13] 何瑾琳, 刘学军, 徐新艳, 毛宇佳. 融合node2vec和深度神经网络的隐式反馈推荐模型[J]. 计算机科学, 2019, 46(6): 41-48.
[14] 赵海燕, 汪静, 陈庆奎, 曹健. 主动学习在推荐系统中的应用[J]. 计算机科学, 2019, 46(11A): 153-158.
[15] 朱峙成, 刘佳玮, 阎少宏. 多标签学习在智能推荐中的研究与应用[J]. 计算机科学, 2019, 46(11A): 189-193.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] 雷丽晖,王静. 可能性测度下的LTL模型检测并行化研究[J]. 计算机科学, 2018, 45(4): 71 -75 .
[2] 孙启,金燕,何琨,徐凌轩. 用于求解混合车辆路径问题的混合进化算法[J]. 计算机科学, 2018, 45(4): 76 -82 .
[3] 张佳男,肖鸣宇. 带权混合支配问题的近似算法研究[J]. 计算机科学, 2018, 45(4): 83 -88 .
[4] 伍建辉,黄中祥,李武,吴健辉,彭鑫,张生. 城市道路建设时序决策的鲁棒优化[J]. 计算机科学, 2018, 45(4): 89 -93 .
[5] 史雯隽,武继刚,罗裕春. 针对移动云计算任务迁移的快速高效调度算法[J]. 计算机科学, 2018, 45(4): 94 -99 .
[6] 周燕萍,业巧林. 基于L1-范数距离的最小二乘对支持向量机[J]. 计算机科学, 2018, 45(4): 100 -105 .
[7] 刘博艺,唐湘滟,程杰仁. 基于多生长时期模板匹配的玉米螟识别方法[J]. 计算机科学, 2018, 45(4): 106 -111 .
[8] 耿海军,施新刚,王之梁,尹霞,尹少平. 基于有向无环图的互联网域内节能路由算法[J]. 计算机科学, 2018, 45(4): 112 -116 .
[9] 崔琼,李建华,王宏,南明莉. 基于节点修复的网络化指挥信息系统弹性分析模型[J]. 计算机科学, 2018, 45(4): 117 -121 .
[10] 王振朝,侯欢欢,连蕊. 抑制CMT中乱序程度的路径优化方案[J]. 计算机科学, 2018, 45(4): 122 -125 .