Hybrid Recommendation Method Based On Heat Conduction And Mass Diffusion

With the large-scale popularization of the Internet and the rapid development of the information age, the amount of network data shows an explosive growth trend, resulting in the problem of information overload. How to get the information fast they really want from the massive data has been a hot research topic. Currently, the recommender system is the best way to solve the problem. Recommendation algorithms are the core of the recommender systems, in which the network-based recommendation algorithm is the focus of this paper. The basic idea is: abstract users and objects as nodes, ignore the basic characteristics of the target users and the recommendation objects, considering only whether there is a relationship between them, and then create the recommendation model to recommend for users by calculating the relationship. The main content of this paper is based on the existing research, studies the network-based recommendation algorithm, heat conduction and mass diffusion.Firstly, the paper summarizes the definition, general model, elements and current research of the recommendation system, and introduces in detail the basic ideas and processes of mainstream recommendation algorithms. Then it studies the heat conduction recommendation algorithm and mass diffusion recommendation algorithm, and analyzes their advantages and disadvantages. Heat conduction recommendation algorithm has good diversity and low accuracy, while mass diffusion recommendation algorithm has good accuracy and low diversity. To solve these problems, two improved hybrid algorithms are proposed in this paper as follows:(1)FHTM algorithm that based on cross hybrid strategy. In order to highlight the impact of users in the whole process of recommendation, resources when transferred are reallocated according to the user's degree in FHTM algorithm. The experimental results show that compared with the heat conduction algorithm, the accuracy of the proposed method is greatly improved, while the diversity is slightly decreased. Compared with the mass diffusion algorithm, the accuracy and diversity are almost identical.(2)For the problems of heat conduction, mass diffusion and FHTM algorithm, this paper proposes a new hybrid recommendation algorithm, HMW, based on heat conduction and mass diffusion. On the basis of the existing research results, this method realizes the personalized recommendation by introducing the user activity to adjust the effects of users in the whole process of recommendation system.Finally, this paper uses standard data sets, MovieLens and Netflix, and the proposed algorithm HMW is made detailed contrastive experiments and analyses with hybrid algorithm HHM on the precision, recall, ranking-score and diversity. The experimental results show that the method HMW is better than the previous algorithm HHM in accuracy and diversity.