关键词: 支持向量机, 统计学习理论, 模式识别, 模糊集

Abstract: A new support vector machine （SVM） algorithm based on multi-classed fuzzy compensation （M-FSVM） is described in more general format to solve the classification training bias problem in the traditional SVM application. And the algorithm is programmed based on the LibSVM code and is used to classify the network intrusion data. The test results show that the small size class classifier precision is improved.

Key words: support vector machines, statistical learning theory, pattern recognition, fuzzy set