A personalized recommendation strategy based on trusted social community

Personalized recommendation technology has become a very effective approach to cope with the problem of information overload in E-commerce. Currently, there are three recommendation strategies: content-based recommender strategy, collaborative filtering recommender strategy, and hybrid strategy. Since characterized as simple, easy to implement, and with high accuracy, collaborative filtering recommender algorithm is widely used. But typical problems still exist in traditional collaborative filtering algorithms, for example, data sparsity, cold start, easy to be attack, and poor ability for migration of user preference, etc, which lead to decreasing of recommendation accuracy and decline of user confidence for recommendation system. In this paper, a novel personalized recommendation strategy is proposed based on traditional collaborative filtering technology, aimed at the issues of poor interest migration and easy to be attack. We call it trusted social network community based recommendation strategy Trusted communities in which users all with similar preference are been detected from user social network. And then recommendation is been given based on these trusted communities. Nodes in the trusted community all have trust relationship between them, so it can effectively avoid attacks from malicious nodes. Meanwhile, a mechanism used for trust relationship feedback has been introduced. Users can submit some feedbacks for their trust relationship after every transaction, which means while the quality of item they received surpass their expectation, they could enhance the relationship, and while fall behind their expectation, they can weaken the relationship. This mechanism provide a good approach to solve change of preference. Experiments show that the algorithm can effectively improve the accuracy of recommendation, and enhance customer satisfaction.