Research On Sina Weibo 's Data Acquisition And Recommendation

As a new social application platform, the development of Microblog in recent years more and more better. Scholars have been doing in-depth research on microblog. In the era of information explosion. People can’t pay attention to all microblog in the massive microblog. So recommendation system appeared. The traditional microblog recommendation using the topology of SNS and user’s registration information to give the recommend result. but the effect is not very good. In this paper, Considering user’s history microblog. According to predict user’s interest to give the recommend result. The result will more properly.In this paper, we do deep research on microblog recommendation system realization process. The core of the work is divided into three parts. The first part is the data acquisition of sina microblog:Mainly through Sina API and web crawler to collect microblog data. The collect result as corpus and the source of recommendation system. The second part is cluster users:analysis all the microblog published by users and extract the user-topic model. Then we use k-means algorithm to divide user into k classes. The user’s interest in same classes are similarity. The third part is the produce of recommendation list: Using all topic word to generate the user feature vector. Then use the text similarity algorithm which based on word2vec to compute the user feature vector and all the other user’s microblog in the same class, the result will return a similarity list. Sorting the similarity list in order, put the Top-N value into recommendation list. In fact, a value is a microblog in recommendation list, recommend result microblog to user.At the end of paper, we use the express (a web application framework for node) to build a microblog recommendation system which realize the function of microblog recommendation.