【摘要】 针对传统的机器学习分类算法在非平衡数据集上准确率较低的问题,提出一种基于随机森林Gini指标和卡方检验的最优特征子集的特征选择方法 (RFG-χ~2),并应用于支持向量机算法模型中。利用SMOTE算法对非平衡数据集进行预处理,通过权衡特征的关联性和特征强度这两个指标,训练生成支持向量机模型所需的最优特征子集。实验结果表明,该方法在支持向量机(SVM)模型上筛选的两个特征子集对应的分类精度分别提高了2.5%和1.5%。更多还原

【Abstract】 A feature selection method based on random forest Gini index(RFG)and chi-square test(χ~2)was presented based on support vector machine(SVM)model to solve the traditional machine learning classification algorithm for the problem of low precision in imbalanced data set.Synthetic minority oversampling technique(SMOTE)algorithm was introduced to preprocess the imbalanced data set,the optimal feature subset was trained to generate SVM models by balancing the two characteristics of feature relevance and feature intensity in the method.Results show that the classification accuracy of the two feature subsets filtered using the method on the support vector machine(SVM)model is increased by 2.5% and 1.5% respectively.更多还原