A Method based on One-class SVM for News Recommendation

Abstract

In order to provide intelligent recommendation and personalized service for users on news website, this paper presents a method based on One-Class SVM for news recommendation algorithm. By analyzing the news webpages and user's browsing history, and by building One-Class SVM model, this algorithm can recommend news for user. The main work of this paper is to study this news recommendation algorithm and to show its experimental results under Dot NET platform. First, this algorithm preprocesses the webpages from Sogou Labs, each of which has its inherent domain and builds One-Class SVM models for these domains. Next, it builds user interest models for each user by analyzing their browsing histories. Then it finds the user's most interested domains by comparing each domain models and user interest model. Finally, it utilizes the webpages of these domains and user's browsing history to build One-Class SVM model to calculate the most relevant webpages to user interest, and recommends these webpages to user. This algorithm takes the lead in calculate the similarity between user interests and webpages using One-Class SVM model and apply hierarchical model to make the results more accurate. From the results, we can find that this algorithm is running pretty well.