Research And Implementation Of Clothing Recommendation Alogrithm Based On Improved Collaborative Filtering

The development of Internet technology and e-commerce sites are flourishing. In recent years, sales volume and saleroom of clothing has been greatly increased. But online shopping sites not only make shopping more convenient, but also force users to experience the burden of information overload. At the same time, users are also faced with lacking of personalized service when buying clothing goods online, In the face of vast amounts of online shopping resoources, the generation of recommendation system is to assist users and operators of online shopping company to ease the problem of information overload. Collaborative filtering algorithm is one of the many recommendation algorithms, and it is also one of the recommended techniques to date, however, the traditional collaborative filtering recommendation algorithm faces many challenges with the vast amounts of network information, such as sparsity of data, the difficulty to measure the similarity of users, the poor scalability, the "cold start" and so on.All of these affect the quality of the recommendation system. In view of these questions, this paper proposes the corresponding improvement as well as the supplement to this algorithm,the main work is as follows:Firstly, we implement the user oriented mogujie data crawler, The web crawler can capture the sharing record of all users and record the user’s list of concerns. Then, the crawled data is utilized to initialize the clothing recommendation system.Secondly, we research the user’s similarity model based on the social relationship information data of consumers. In this paper,we also separately from the users to share in the apparel commodity text and image contents two aspects to construct the similarity between users based on text and image classification technology for users’sharing of the apparel commodity information; For the user’s social network data,considering the mutual concern information between users, we integrate the apparel goods text content and visual content, the user’s network information through user similarity function.Thirdly, we propose a similarity model based on the user’s social relationship data and user-item rating matrix data, and propose a collaborative filtering recommendation algorithm which is integreated with the social netwrok information. As the experimental result shows, the sparsity of data has been alleviated and the final recommendation of the procudt is more satisfactory because of the intetration of the user’s social network information.Fourth, we realize the online clothing recommendation system. After logging in the system, users can access to all users’sharing information of clothing in the dataset and the attention of information between users. What’s more, the recommended results can be graded by everyone.