Research And Application Of Distributed Support Vector Machine Based On Hadoop

With the development of Internet technology, it is an important issue about how to extract valuable information from the exploding data on the network. So that data mining has got great development in recent years.Support vector machine (SVM) is a highly effective and reliable data classification algorithm in data mining field, which has outstanding advantages in solving nonlinear and high dimensional pattern recognition problems. However, the traditional SVM is not suitable for mass data processing, mainly because of the large datasets increase the SVM train time significantly, slow training speed, and increase the cost of getting model. In order to solve these problems, it is an effective method to distribute parallel the traditional SVM algorithm.In this paper, the distributed parallelization of SVM algorithm is discussed. The existing four kinds of support vector machine strategies of grouping, cascade, feedback, mixed parallel are analysised from basic idea, training effect, implementation difficulty. After comparison, as a way of realizing distributed SVM, the feedback parallel support vector machine (FeedbackPSVM) can effectively reduce the training time, ensure accrucy and easy to understand and realize.In this paper, we design and implement a FeedbackSVM algorithm based on distributed storage and parallel computing of Hadoop which is used to diagnose low voltage. The experimental results show that the parallel SVM algorithm’s training time is significantly less than the single SVM algorithm, and the accuracy can be essentially flat, so it has very high research significance and practical value.