Data Processing And User Behavior Analysis Based On Sina Weibo

With the continuous development of Internet technology, social network is playing a more and more important role in people’s daily life, meanwhile changing the way of information spreading, from original print media and television broadcast to the combination of diversified platform. The relationship between social network and the way that people get popular information and their information of interest is more and more inseparable. Accompanied by social network, people is making deep mining on user’s behavior through the analysis of a large number of user’s data, in order to save people’s time on valuable information extracted from huge amounts of information and mine the potential business value. The success of Facebook and Twitter largely promotes the development of the domestic social network platform.We study the information based on the domestic popular social network platform, Sina Weibo, and mainly complete the following four aspects.First, we study the relatively popular web crawler technology. Through comparison and comprehensive analysis, we design and implement a data acquisition system which can acquire and store weibo content, user information, user relationship information, weibo relation information to extract features on our demand. We also design the corresponding database according to the data correlation. Besides, we not only use multi-thread technology to dramatically improve the efficiency of the crawlers, but also design the multi-user authorization mechanism to break the API restrictions rules of Sina and ensure the uninterrupted work of the crawlers.Second, we study 13 features from Sina Weibo through machine learning and set up a feature analysis model to predict user’s forwarding behavior and find out the weibo features which are important to this behavior. And we study the feature combination at the first time to find out the correlation between features. Meanwhile, we train models with different machine learning algorithms to find out the best one.Third, we set up a Hybrid Classifier Sentiment Prediction Model(HCSPM). This model extracts the prediction results of each classifier and weights them, through four kinds of machine learning classification algorithm, to predict the weibo emotion classification.Fourth, we design a user behavior analysis system to improve the efficiency of research in the process of experiment and make the analysis of experiment results more intuitive and accurate, which combines different classification algorithms and make the analysis of user behavior more clear and efficiency. Above all, in this paper, we analyze user’s behavior based on Sina Weibo and display the main direction of further research.