Latent Dirichlet Allocation(LDA) model can be used for identifying topic information from large-scale document set, but the effect is not ideal for short text such as microblog. This paper proposes a microblog user model based on LDA, which divides microblog based on user and represents each user with their posted microbolgs. Thus, the standard three layers in LDA model by document-topic-word becomes a user model by user-topic-word. The model is applied to user recommendation. Experiment on real data set shows that the new provided method has a better effect. With a proper topic number, the performance is improved by nearly 10%.

For the widespread use of microblog business and the impact on data mining techniques, a mining algorithm of microblog interpersonal relationship network is proposed based on the fuzzy matching of tag, and the characteristics of the network are analyzed. Use the tag of the users, the algorithm mainly considers word morpheme, order, and word length to calculate the match degree of the words when matching the tag. For weakening the influence that using different users as a starting point may have different result, ordinary users and celebrities as a starting point separately are used. At the same time, the structural characteristics of the network are studied, and the analysis results show that the network has small-world and scale-free properties. The results show that the mining error rate of celebrities and common users friends who are interested in IT. When mining 10 celebrity users’ friends, the average error rate of the algorithm is 14.08%, and 10.63% for common users.

Currently, Web crawler and microblog API which are used to grab data from the microblog are difficult to satisfy the public opinion system demands for microblog data. To settle the problem, this paper presents a feasible solution which is the similar as the browser login microblog to capture data from Web pages. It can easily get all data from any microblog users. On this basis, it constructs a microblogging network through interconnections among users, and discovers new users through it. In order to get high quality data, it builds mathematical models to calculate the user’s influence index by using posting number, posting frequency, fans number, forwarding number and comments number. Moreover, it builds priority queue according to the calculated influence factor, which let those that have bigger influence index have high acquisition frequency. Finally, it calculates time interval to balance the lower frequency of non-active microblog user. The experimental results show that this method not only processes easily and has higher speed but also can obtain high quality information and have huge versatility.

In order to solve the problem of a large number of advertisements on Sina, Tencent microblog platform, this paper proposes a microblog advertisement filtering model. Through the data pretreatment, the raw data are converted into clean data and easy to be handled by the computer. In the pretreatment stage, according to the characteristics of the microblog, this paper emphatically improves the stop word list, and it plays a key role in improving precision. Then it builds a classifier based on support vector machine for training data, and through continuous learning and feedback, better classification results are achieved. Experimental results show that the model of advertisement filter achieves better effect, when filtering accuracy is more than 90%, which is better than the method based on keywords.