Research On Collaborative Topic Regression Recommender Algorithms Based On SocialNetwork In Party-construction Field

With the rapid development of information technology, e-commerce and social media appeared in people’s vision. Internet quickly changing people’s life style, but also change the party’s environmental governance. Therefore, it is a new task of our party which is enormous significance to extend party-construction work on Internet and service better for Party-mass. Keeping up with the pace of the party, the construction of grassroots integrated service platform was started within the scope of Yunnan, but with the increasing of party members and party-construction information which posted on the microblogging platform of party-construction, there is a problem same as other social media, namely the explosive growth of information brought by information overload. As we all known, there are two ways to mitigation information overload problem, one is the search engine, the other one is recommendation system. Therefore, the paper will propose an improved algorithm based on the study of recommended methods, trying to ease the information overload problem, specific for the following discussion and research:(1) The analysis of Yunnan grassroots integrated service platformThis paper analyzes the function of the Yunnan grassroots integrated service platform, and research the information overload encountered in the microblogging platform of party-construction, note that, we also need to use recommendation algorithm to reduce resource consumption, saving time costs.(2)Proposed a term Social Regularization based on social trustSimply introduces the social trust relationship and traditional recommend methods. We interpret the relationships between users and propose a term Social Regularization to represent the social constraints on recommender systems(3)Collaborative topic regression function with social regularizationBased on the study of the social network, we propose our CTR-SR model and systematically illustrate how to design Collaborative topic regression function with social regularization, how to learn parameters and how to predict rates using the new function. We use Last.fm dataset and microblogging dataset of party-construction to comparing CTR-SR with CTR, LACTR, CTRSMF using different quality indicators of item recommendation, such as the accuracy, diversity, novelty. The experimental results show that CTR-SR achieves the best accuracy in both precision and recall when using the Last.fm dataset. When using the microblogging dataset of party-construction, CTR-SR is better than LACTR, CTRSMF, but lower than LACTR.