Design And Implement Of Collaborative Filtering Algorithms Based On Web Log And Clustering Analysis

In recent years, e-commerce has been rapid development, more and more e-commerce systems for users with more choices at the same time, so that users are often lost in a large number of commodity information, it is difficult to quickly find themselves in need of goods. E-commerce recommendation system directly with the user interaction, simulated shop sales workers to provide users with recommendations of goods, to help users find the products, thus the successful completion of the purchase process. As e-commerce system to further expand the scale of e-commerce recommendation system is also facing a range of issues, such as: data sparse, system scalability, cold start and other issues, address these issues, the paper recommended method of e-commerce a useful Exploration and research, and to the corresponding solutions.The traditional method of commodity users in the evaluation of the sparsity of information and the resulting low quality of the recommendation of the problem and the server log analysis method, the user browse or buy merchandise in the process shown by the degree of interest in the hidden goods According to certain information algorithm to translate them into the dominant users of the commodity numerical score, the more evaluation information as collaborative filtering algorithms based on the data, the accuracy of the algorithm were recommended raising played a big role.The traditional algorithms because of the increased volume of data and online data processing algorithm led to the problem of low efficiency, this paper presents the user clustering method will be of similar interest to the users of the same cluster, the work can be carried out off-line, When a target user to recommend only need and the similarity of its interest in the search for their neighbours in the cluster, then neighbours interest to the target users are most likely to recommend the products that interest, this method can be significant savings online Algorithm data processing time, thereby enhancing the collaborative filtering algorithm efficiency of the recommendation.Finally, the improved algorithm proposed in this paper the experimental verification, in accordance with the experimental data obtained improved algorithm is reasonable and effective.