Abstract:In this era of information explosion, how to handle these vast amounts of data and how to classify the data effectively has attracted much attention, especially in the stage of rapid development of Internet technology free, the field of web classification has become a hot spot. Compared with the traditional classification methods, support vector machine has the characters of high-dimensional, small sample size, strong adaptability, and can be very effective to solve the problem of web page classification. But in the field of classification of imbalanced data, there is a problem of inaccurate classification. Therefore, this paper proposes a new strategy to solve the imbalance data samples, that is, combining the under-sampling strategy with the traditional support vector machines to increase the number of samples set in the minority class and to reduce the concentrated noise data in the majority class, so that imbalanced sample set tends to be balanced. Finally SMO algorithm is used to improve the accuracy of classification.