期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2017
卷号：6
期号：1
页码：73-76
出版社：Shri Pannalal Research Institute of Technolgy
摘要：In recent years, shopping online is becoming moreand more popular. When it need to decide whether to purchase aproduct or not on line, the opinions of others become important. Itpresents a great opportunity to share our viewpoints for variousproducts purchase. However, people face the informationoverloading problem. How to mine valuable information fromreviews to understand a user’s preferences and make an accuraterecommendation is crucial. Traditional recommender systemsconsider some factors, such as user’s purchase records, productcategory, and geographic location. In this work, it propose asentiment-based rating prediction method to improve predictionaccuracy in recommender systems. Firstly, it propose a social usersentimental measurement approach and calculate each user’ssentiment on items. Secondly, it not only consider a user’s ownsentimental attributes but also take interpersonal sentimentalinfluence into consideration. Then, consider item reputation, whichcan be inferred by the sentimental distributions of a user set thatreflect customers’ comprehensive evaluation. At last, by fusingthree factors-user sentiment similarity, interpersonal sentimentalinfluence, and item’s reputation similarity into recommendersystem to make an accurate rating prediction. It conduct aperformance evaluation of the three sentimental factors on areal-world dataset. Experimental results show the sentiment canwell characterize user preferences, which help to improve therecommendation performance.
关键词：Item reputation; Reviews; Rating prediction;Recommender system; Sentiment influence; User sentiment.
