The Research Of Content-Based And Collaborative Filtering Algorithm In Recommendation System

As the E-commerce is more and more convenient, the backend structure is more complex. Users have a greater choice of space,but at the same time, the appearance of mass information also hinders users to find what they want quickly and accurately.Recommendation system is born to solve such situation, it can be customized for users to find what they want. Today, almost all e-commerce websites have applied recommended system, such as Taobao, ePay, etc.. In order to produce a higher quality of recommendation, technical personnel have proposed a variety of recommendation methods, including content-based recommendation, collaborative filtering recommendation, association mining recommendation, etc.. However, with the development of the research and the improvement of users’ requirements, the recommendation algorithm has gradually exposed some shortcomings, such as cold start and data sparsity problem of collaborative filtering, and the balance between the quality and the real-time performance.This paper mainly focuses on the content based algorithm and collaborative filtering algorithm. Here’s the detail:(1) It has a summarization of recommendation algorithm’s development process and key technology.(2)By offline pre-processing data to create virtual space, the performance of online realtime recommendation is tuning and the issue is solved.(3) A hybrid algorithm is proposed, which is based on the collaborative filtering algorithm and content based algorithm. The new hybrid algorithm combines user preference and project features into user preference model, and use it to do K-means clusters,which finally makes recommendation. The improved algorithm not only can solve the problem of data sparsity of the traditional collaborative filtering algorithm, but also can predict the potential users of interest in new projects. What’s more, it can also solve cold start issues effectively.Experiments show that the improved algorithm can solve the problem of speed bottleneck, which is originally caused by data sparsity, cold start and online recommendation. And it can also improve the quality of recommendation.