Research On Social Network Recommendation Algorithm Based On Multi-source Information

In the era of Web2.0,people's daily life is inseparable from the Internet.The user's passive reception of information to the active acquisition of information makes the amount of data increase and the data types complex and diverse.Because users can't quickly and accurately locate the information which they need from massive information and produced the phenomenon of "information overload".The "information overload" problem exists in the recommendation system has two main manifestations are data sparsity problem and data scalability problem.Collaborative filtering is an information filtering technology that can solve the problem of poor current recommendation performance.Therefore,in order to solve data sparseness problem and scalability problem of the recommendation system in social network environment,this thesis uses collaborative filtering as the basis and integrates social network information to conduct algorithm research.The main work and innovations of this thesis are as follows:1.Aiming at the problem of data sparsity in the social network recommendation system,the probability matrix decomposition method is used as the recommendation framework.In order to realize the design of collaborative filtering recommendation algorithm,the four parts of user project feature attribute,project relevance degree,user personal feature attribute and social feature attribute are combined in the same framework by the form of matrix decomposition.The effectiveness and feasibility of this recommendation system framework for social network environment is verified.The experimental data simulation analysis shows that the proposed algorithm can effectively solve the data sparsity problem in the recommendation system and reduce the average absolute error and root mean square error of the system.2.The big data environment has the characteristics of large data volume,fast data update and low data utilization rate.It is an effective method to improve the data utilization efficiency by labeling the data in advance by using tags.Therefore,in view of the scalability problem existing in social network recommendation system,based on collaborative filtering recommendation and combined with the smart community service platform,the recommendation problem is transformed into classification problem for research.This part introduces the data pre-processing scheme in first;Secondly,the topic community is mined by the joint NMF model and the K-means clustering idea is applied to the classification to generate the topic community;Finally,in the process of generating Top-N recommendation list,the two-degree separation update mechanism combining loss function and Kullback-Leibler divergence is perform data dynamic update to find the optimal solution.The experimental results show that the proposed algorithm can effectively solve the problem of data scalability in the recommendation system and improve the accuracy and recall rate of the system.