一种改进的分类算法在不良信息过滤中的应用

使用KNN(K Nearest Neighbor)分类算法进行不良文本信息过滤时,由于包含不良信息的样本不易获取,导致分类器预测结果严重倾向于多数类。为改善少数类过滤效果,从数据层的角度改进了传统的KNN算法:先将少数类样本聚类分组,再在每个聚类内部使用遗传交叉生成新样本,并验证其有效性,最终获取到各类别样本数量基本均衡的训练样本集合并训练KNN分类器。实验结果表明,本文的方法可有效识别不良文本。此方法同时适用于其他关注少数类分类精度的不均衡数据集分类问题。

Application of a Improved Categorization Algorithm in the Malicious Information Filtering

The prediction result of classifier was biased towards the class with more samples,because of the samples that including the malicious information were difficult to gain when using the KNN(K Nearest Neighbor) categorization algorithm to filter the malicious information.In order to improve the filter effect of the class with fewer samples,the traditional KNN algorithm was improved from the data angle: the class with fewer samples was grouped by using cluster algorithm,then the genetic crossover operator was ...