Abstract: SVM is an effective classification method，the nature of learning is that it solves the original problem by the dual problem，but it does not directly obtain the feature importance.This article raises a new feature selection algorithm.Experiments show that the new feature selection algorithm and F-Score algorithm for calculating the same test data set is the result of the same descending order.The new feature selection algorithm is reasonable.

Key words: Support Vector Machine（SVM）, feature selection, F-Score

摘要： 支持向量机（Support Vector Machine，SVM）是一种有效的分类方法，其学习本质是通过对偶问题求解原问题，但是它不能直接获得特征重要性。提出一种新的特征选取算法，实验表明，该特征选取算法与一般特征选取算法（如F-Score算法）相比，对同一测试数据集计算的结果具有相同的降序排列结果，而且有更好的特征刻画量化指标，分界线更明显，表明新的特征选取算法具有更佳的合理性。

关键词: 支持向量机, 特征选取, F-Score