Research On Microblog Following Recommendation Method Based On Spectral Clustering

Micro-blog has become a major platform for internet users to access information,express ideas, and communicate for each other, due to its fast spreading, stronginteractive, and free accessibility, the growth of the information is very fast, the vastamounts of information make it difficult for the users to find their recommendedtarget which they are interested in. In this context, we need a new personalizedrecommendation service for different micro-blog users, which can recommend itemsthat meet their interests preferences based on the characteristics and interests of users.The personal recommendation can not only improve the performance of micro-blogmarketing, but can also improve the users’experience. How accurately advertising hasbecome one of the enterprise micro-blogging marketing issues to be addressed.This paper aims to propose improvements to the current concern micro-blogfollowing recommendation method design ideas, based on the construction ofmicro-bloggers’ interest model, we could mining the relevance of the micro-bloggers’following objects by the spectral clustering method supplemented which combines thecollaborative filtering recommendation, focused on solving the following twoquestions:First, the micro-blog user interest model based on multi-dimensionalcharacteristics of the construction. Constructed based on a single dimension for thecurrent user interest model micro-blog cause concern recommendation effectiveness isnot high status, proposed micro-blog user interest model based on multi-dimensionalcharacteristics of the construction method, it determines the source of the micro-bloguses by analyzing the micro-blog user characteristics and behavior, and converts itinto user interest, while select micro-blog user interest features and on this basis,constructs the user interest model from three dimensions: the micro-blog userattributes, keywords and user behavior.Second, the design is based on implicit rating micro-blog followingrecommendation method. Concerned about the lack of current recommended methods,proposed a new micro-blog following recommendation method based on user attributes and implicit rating, designed implicit rating from the micro-blog userbehavior (@, forward, comment), and calculated the similarity between users. On thisbasis, to join micro-blog user attribute information to calculate the similarity betweenusers, and finally filter neighbor by the user and to weighted combination of both, bythis method, it can improve the collaborative filtering algorithm and traditionalneighbor user selection method at the same time, so that the recommendations to thetarget users can be more precise.Experimental results show that the spectral clustering micro-blog followingrecommendation method is more accurate and effective than traditional collaborativefiltering technology. Based on understanding the preferences of the micro-blog users,this method analyze deeply of the micro-blog users’ characteristics and behavior,improve the traditional recommended method. It not only provides guidance for theenterprise micro-blog marketing strategy, but also provides accurate micro-blog usersefficient micro-blog following recommendation.