基于提升度的KNN分类子的分类原则改良模型

Improving Classification Rule with Lift Measure for KNN Classifier

针对非均匀类数据,本文提出K最近邻分类子的一个分类原则改良方法,能够度量待分类数据的K个近邻点中的类比率提升量,增大了最小类数据的竞争力,明显地提高了小类数据的分类正确率。实验结果表明,本文提出的改良分类原则对非均匀数据分类的准确率明显高于传统的KNN分类算法。

A KNN classifier is presented for classifying imbalanced data.A gain model is constructed for measuring the lift of probability of a class label.The competition of minority class is well enhanced in imbalanced-class dataset.And the accurate rate of classifying minor-class data is significantly improved.The experimental results show that in the setting of imbalanced-class datasets,the proposed approach has significantly improved the classification accuracy,compared with the existing KNN classifiers.