Research on microblog filtering technology based on improved online support vector machine model

With the rapid growth of the number of micro-blog, a lot of useless information flooded in the users vision, the users find it difficult to choose micro-blog recommendation service according to personal interest, so micro-blog filtering technology is applied to the micro-blog service. The interests of users change with the time, so the traditional batch learning can not be able to satisfy the need of users interest model. However, the machine learning which is based on online learning solved these problem in a certain way. The optimization of micro-blog filtering can decrease classification errors and improve the classification of data, but it has its own disadvantages. Although online support vector machine filtration is excellent, there's long-running shortcomings. The paper focus on the improvement of efficiency of the online support vector machine filtering by reducing the size of the training set, the number of training and the number of iterations. The experimental results show that the filtering performance fluctuate slightly, but it can almost be ignored because of the advantage of its efficiency, and the great amount of data is, the greater the efficiency becomes more obvious.