摘要： 将机器学习中的经典分类方法与规则方法相结合，用以分析新闻语音文本的情感倾向，并判断其强弱。通过支持向量机分类器来研究特征选择方法及特征权重计算方法的组合对实验结果的影响。在实验过程中发现适当的结合规则后，实验结果在不同程度上都有了提高，在KNN和Bayes分类器上做了对比实验，结果证实该方法的普适性。

关键词: 机器学习, 特征选择, 特征权重, 支持向量机

Abstract: This paper uses machine learning techniques combined with the rules to solve sentiment classification of news text, and researches affection of feature selection and feature weights based on Support Vector Machine(SVM) classifier. Experiments show that combined with the rules, experimental result is improved. In order to test universality of the combined method, more experiments based on KNN and Bayes classifier are done. Results show that combined method does better than not combined ones.

Key words: machine learning, feature selection, feature weight, Support Vector Machine(SVM)

中图分类号: 

• TP391.1