Research And Application Of Commodity Recommendation Algorithms Based On Clustering Methods

With the development of e-commerce sites, more and more researchers begin to study recommendation systems. Recommendation system can recommend the commodities and information which users are interested in to users according to the interests and purchase history of the users. Recommendation systems can take the initiative to mine the potential users who have the possibility to buy some certain product. Similarly, it also can help users to find what they are interested in. By this way, it can increase sales of the e-commerce website and also increase users’loyalty.Firstly, this paper has fully studied the research background and research progress at home and abroad of recommendation systems. Currently, e-commerce applications are developing rapidly and more and more, recommendation systems are applied into e-commerce websites. Although recommendation system has been studied by researchers for more than ten years, the recommended accuracy of recommendation systems still cannot reach the desired effect and how to increase the recommended accuracy is still the hot issue of recommendation system research.On the basis of research studies, this paper introduced some conceptions about recommendation algorithms and explained the fundamental of recommendation algorithms. After that, it introduced the advantages and disadvantages of two recommendation algorithms:the content-based recommendation algorithm and the collaborative filtering recommendation algorithm. For the content-based recommendation algorithm, the biggest problem is that it is difficult to describe some certain product, and for the collaborative filtering recommendation algorithm, the most problem is called data sparsity. In order to study how to resolve the data sparsity and how to increase the recommended accuracy, this paper introduced and studied the relevant techniques of clustering algorithms.In order to resolve data sparsity existed in collaborative filtering algorithm, this paper proposed an improved collaborative filtering recommendation algorithm which is based the method of evaluation filling and introduced k-means algorithm to further increase the evaluation accuracy.Finally, this paper used Movielens which is a data set of real ratings as the research objects and used matlab technology to verify the proposed clustering-based collaborative filtering recommendation algorithm. It mainly studied the effects of matrix density, k value of clustering method and the neighbor number on recommendation results. Finally, the proposed algorithm are evaluated from MAE、recall、precision and F1. Experimental results show that the proposed collaborative filtering recommendation algorithm based on clustering method has a higher accuracy than the traditional recommendation collaborative filtering recommendation algorithm.