Hybrid Recommendation Algorithm Based On Networking Feature

With the improvement of computing ability and storage capacity, we are entering the area of DT.The information has been growing in an explosive way and the cost of finding the content is increasing, so people fall into serious information overload. Recommendation system which is an important form of solving the information overload, has been developed rapidly in recent years. Especially in the field of electronic business, recommendation system has been widely used, recommendation system can solve the marketing problem of long tail goods, and at the same time, it can provide users with personalized recommendations, which could improve the user experience. Recommendation system played an important role in the development of the electronic business.The traditional collaborative filtering algorithm is certainly very mature recommendation algorithm. Firstly, collaborative filtering is a highly cost-effective recommendation algorithm, which requires minimal domain knowledge to build a model and does not require complex feature works. It embodies the idea of collective intelligence. Secondly, it can solve the recommended problem of the non interactive products. In the case of sufficient user’s behavior data, the advantage of collaborative filtering algorithm is more obvious.However, collaborative filtering is also facing some problem. Firstly, it has cold-start problem. The lack of needed behavioral data when new user and item join the system can seriously hurt the performance of the recommendation system. Secondly, it can not solve the problem of sparse data, which will affect the effectiveness of the algorithm. In addition, it can not add more features to describe item and user, if we just improve the similarity function, which can not significantly improve the effectiveness of the algorithm.In the field of electronic business, the most important role of recommendation system is to improve the user experience and increase the volume of goods, however, with the expansion of electronic business and the rapid growth of data, the traditional recommendation algorithm can not meet the demand. So this paper adopts the recommendation method which is based on the match and rank, the hybrid recommendation method enables us to use more features to capture user’s preference. In the aspect of improving the recommendation accuracy, this paper applies the complex network algorithm to mine the relationship features of user’s behavior, and extends the feature of this dimension into the model. In addition, the community discovery algorithm used in this paper is distributed implementation, which makes it possible to mine large-scale relational data.In the last chapter of the paper, it is found that the proposed hybrid recommendation algorithm can significantly improve the effectiveness of the recommendation model.