Paper Info
Title
Pairwise Preference Over Mixed-Type Item-Sets Based Bayesian Personalized Ranking for Collaborative Filtering
Abstract
Nowadays, providing high quality recommendation services to users is an essential component in online Web applications, including shopping, making friends, healthcare, etc. In some recent works, the recommendation problem of one-class collaborative filtering has been proposed and been receiving more attention. This approach can make use of users' behaviors with the "one-class" feedbacks form coming from different services to improve the recommendation performance, e.g., "purchasing" in shopping services, "watching" in video websites, and "making appointment" in online health community. Previous works solve the issue via proposing the assumptions of pointwise preference on one item, pairwise preference on items or item-sets based on relative score over two item sets, in which one previous work named Bayesian personalized ranking (BPR) was empirically found performing much better by means of leveraging such one-class data well. Nevertheless, such pairwise preference assumptions with regard to items or item-sets may always invalid. In this paper, we present an improved assumption of pairwise preferences over mixed-type item-sets by defining the preference on two item sets with different type instead of on a simple item set. One is an item set with the same-type feedback relationship; the other is a mixed-type item set. With this improved assumption, we then develop a novel algorithm, MT-BPR (Bayesian Personalized Ranking over Mixed-Type Item-sets). The empirical results based on two data sets, which are collected from a healthcare website and mobile e-commerce application in real world, demonstrate that MT-BPR outperforms than several previous methods in ranking predictions.
Year
DOI
Venue
2017
10.1109/DASC-PICom-DataCom-CyberSciTec.2017.22
2017 IEEE 15th Intl Conf on Dependable, Autonomic and Secure Computing, 15th Intl Conf on Pervasive Intelligence and Computing, 3rd Intl Conf on Big Data Intelligence and Computing and Cyber Science and Technology Congress(DASC/PiCom/DataCom/CyberSciTech)
Keywords
Field
DocType
recommender systems,personalized ranking,implicit feedback,matrix factorization
Recommender system,Pairwise comparison,Collaborative filtering,Ranking,Information retrieval,Computer science,Purchasing,Web application,Bayesian probability,Pointwise
Conference
ISBN
Citations 
PageRank 
978-1-5386-1957-5
0
0.34
References 
Authors
26
6
Name
Order
Citations
PageRank
Shan Gao1136.31
Guibing Guo258634.18
Yusong Lin312.04
Xingjin Zhang400.68
Yongpeng Liu521.71
Zongmin Wang61911.75