AE-MCCF: An Autoencoder-Based Multi-criteria Recommendation Algorithm

被引:28
作者
Batmaz, Zeynep [1 ]
Kaleli, Cihan [1 ]
机构
[1] Eskisehir Tech Univ, Comp Engn Dept, Eskisehir, Turkey
关键词
Multi-criteria; Collaborative filtering; Deep learning; Autoencoders; Accuracy; SYSTEM;
D O I
10.1007/s13369-019-03946-z
中图分类号
O [数理科学和化学]; P [天文学、地球科学]; Q [生物科学]; N [自然科学总论];
学科分类号
07 ; 0710 ; 09 ;
摘要
Recommender systems enable users to deal with the information overload problem by serving personalized predictions. Traditional recommendation techniques produce referrals for users by considering their overall opinions over items. On the other hand, users may consider several criteria while evaluating an item. Even though overall rating-based evaluations are sufficient for interpreting e-commerce products, they may be less effective for evaluating services provided by hotels, restaurants, etc. Accordingly, multi-criteria-based collaborative filtering systems are introduced to increase the level of personalization. These recommender systems are relatively new extension of traditional collaborative filtering systems, and they utilize multi-criteria-based user preferences provided by individuals considering several aspects of services. There are several studies related to such recommender systems, and according to their results, it is possible to produce more personalized predictions along with accuracy improvement by employing multi-criteria recommender systems. Deep learning techniques employed in many research areas such as pattern recognition and image processing have recently been used frequently in the field of recommender systems. The studies show that deep learning-based approaches can improve the accuracy of the referrals due to their high capability of extracting out nonlinear relations between users and items. Therefore, in order to nonlinearly represent relations among users in terms of multi-criteria preferences, we propose a new multi-criteria collaborative filtering algorithm based on autoencoders. Our empirical results show that the proposed method enhances accuracy levels of produced predictions compared with the state-of-the-art algorithms.
引用
收藏
页码:9235 / 9247
页数:13
相关论文
共 40 条
[1]   New recommendation techniques for multicriteria rating systems [J].
Adoinavicius, Gediminas ;
Kwon, YoungOk .
IEEE INTELLIGENT SYSTEMS, 2007, 22 (03) :48-55
[2]  
Adomavicius G, 2011, RECOMMENDER SYSTEMS HANDBOOK, P769, DOI 10.1007/978-0-387-85820-3_24
[3]  
Agathokleous M, 2014, COMM COM INF SC, V459, P205
[4]  
[Anonymous], ARXIV15110728 CORR
[5]  
[Anonymous], 2007, P 24 INT C MACHINE L
[6]  
[Anonymous], 2015, NIPS WORKSH MACH LEA
[7]  
Ashraf M, 2018, INT J SCI RES SCI TE, V4, P61, DOI [10.32628/IJSRST1738526, DOI 10.32628/IJSRST1738526]
[8]   A review on deep learning for recommender systems: challenges and remedies [J].
Batmaz, Zeynep ;
Yurekli, Ali ;
Bilge, Alper ;
Kaleli, Cihan .
ARTIFICIAL INTELLIGENCE REVIEW, 2019, 52 (01) :1-37
[9]  
Bilge A, 2014, INT JOINT CONF COMP, P18, DOI 10.1109/JCSSE.2014.6841835
[10]   On Deep Learning for Trust-Aware Recommendations in Social Networks [J].
Deng, Shuiguang ;
Huang, Longtao ;
Xu, Guandong ;
Wu, Xindong ;
Wu, Zhaohui .
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2017, 28 (05) :1164-1177