The Research And Implementation Of Personalized Recommendation System Based On Web Comments

With the rapid development of Internet, the Internet has become a major source of ourinformation. In all aspects of life, people are closely linked with the Internet. The Internetbrings a fast and convenient life to people, but also a series of problems, the Internet floodedwith overloaded data, make people lost in the world of Internet. Therefore, how should theusers find the valuable information from the overloaded information and let users access touseful information, this hot issues have been involved in academia and business.Accompanied by the BBS, Micro-blog and chat room, the Internet has had a largenumber of web comments. The web comment is the user of the product description, whichcontains a lot of useful information. Sentiment information is extract form the web comments,and it application in personalized recommendation will get better results. But now it is stillvery rare to apply web comments to personalized recommendation.At the same time, personalized recommendation system is rapid, which construct theuse-product bipartite network based on user past activities, and it can recommend usersinformation that they may care based on the similarity with the other users, the popularrecommend technology is collaborative filtering and hybrid recommend algorithm. Thedataset of current recommend technology is come from the relationship of users and products,it’s still very rare to apply web comments to personalized recommendation. This paper willresearch the personalized recommendation system based on web comments, which use thedataset of sentiment information to improve the results of recommendation.This paper research and implement the personalized recommendation system based onweb comments, the main contents are as follows:1. Structural models of the personalized recommendation system based on webcomments use the web comments dataset of restaurant from the public commentwebsite:“www.diaoping.com/”. Firstly, according to the traditional recommendationtechniques, structural the use-product bipartite network, then, structural theproduct-tag bipartite network with the dataset of sentiment information which extractfrom web comments. Finally, we get the personalized recommend model from theuse-product bipartite network and product-tag bipartite network.2. Improve the personalized recommend model. Firstly, a weighted tripartite networkbased algorithm has been proposed, which will improve the recommend results. Thenuse a parameter to regulation product’s degrees, this will reduce the negative effectfrom the product with large degree. 3. This paper will apply the personalized recommendation system based on webcomments to the restaurant recommend from the public comment website“www.dianping.com/”. The experimental results show that the personalizedrecommendation system based on web comments is obvious improvement in rankingaccuracy metric, precision, recall, F1and novelty. So use the web comments data topersonalized recommendation effectiveness.