Research On Collaborative Filtering Recommendation Methods Based On Online Word Of Mouth And User/Item Networks For Network Communities

With the development of e-commerce,some related online communities have also emerged.The online community has a large number of users with the same interests.A lot of users that always write or read the product reviews are common but more professional on the community websites.However,the huge amount of information in the online community also makes each user face the problem of information overload.Current popular recommendation algorithms usually depend on the data of users’ product rating.However,as the size of the website expands,the scoring data are sparser,which greatly reduces the performance of the recommender systems.The online community recommender systems provide more solutions for this,such as by referring to online word-of-mouth data,user social networks,item networks,etc.,and this will improve the performance of the recommendation systems.How to combine heterogeneous information such as consumer preferences,product features,online word-of-mouth,user social networks,and item networks to recommend users valuable information from huge information resources is an issue that the online community recommendation systems need to solve.In view of this,it is necessary to explore online community recommendation methods based on online word of mouth data and user/item networks.The main research works include the following aspects:(1)Proposed an improved collaborative filtering recommendation method based on user word of mouth and social information.Firstly,based on the user’s online comment information and topic mining method,extract the words that represent the user’s preferences in the user’s online comments.Meanwhile convert the unstructured data into structured information and build user similarity measurement model.Moreover,by combining the user’s social information,this paper proposed an improved score prediction method based on user word-of-mouth and user social networks.(2)Proposed an improved collaborative filtering recommendation method based on item word of mouth and item networks.Starting from the items online reviews and using the text topic mining method,extract the words representing the item features in the item comments.Meanwhile,convert the unstructured comment data into structured information,and build item similarity measurement model.In addition,combined with the information of item category and construct item networks,this paper proposed an improved score prediction method based on the item word of mouth and item networks.(3)Showed the application research of collaborative filtering recommendation methods based on online word-of-mouth and user/item networks.To verify the accuracy and practicability of the method this paper proposed,this paper takes the Douban movie data as an example.The experimental results show that using data of online reviews can greatly improve the accuracy of the scoring prediction.In addition,by combining the user/item networks,the data sparseness and cold start problems can be effectively alleviated to improve the performance of the recommendation systems.