摘要
电子商务产品的评论信息对于电子商务产品质量舆情监测具有极大的参考价值.针对集成学习算法在高维度下分类精度降低的不足之处,提出了一种IG-RS-SVM(Information Gain-Random Subspace-Support Vector Machine)算法.以Random Subspace集成学习算法为基础,以支持向量机算法为基学习器.引入了信息增益特征选择算法.通过对特征空间中每个特征的信息增益值进行排序,剔除无价值的特征,降低RS集成算法生成的特征子空间的维度,从而提高了SVM分类算法的效率.实验结果表明,改进后算法可以有效提高评论内容的分类精度.
The remarks of E-commerce products have a great reference value for the public opinion monitoring of E-commerce product quality. Aiming at the deficiency of reducing the classification accuracy with ensemble learning for high-dimensional datasets, a new algorithm, IG-RS-SVM, was proposed. It was based on Random Subspaee, taking SVM as a base learner, and applying the information gain algorithm. By sorting the information gain value of each feature in the feature space, excluding worthless features, and reducing the dimension of feature subspace generated by the Random Subspace algorithm, the efficiency of the SVM classification algorithm was increased. The experimental result shows that the improved algorithm can effectively improve the classification accuracy of remarks.
出处
《中国计量学院学报》
2015年第3期285-290,共6页
Journal of China Jiliang University
基金
国家自然科学基金重大专项项目(No.61027005)
关键词
产品评论
信息增益
随机子空间
支持向量机
production review
information gain
random subspace
support vector machine