摘要
针对乳腺癌智能诊断中的分类器欠稳定,样本分布适应性差等问题。本文提出一种基于Adaboost集成BP、RBF及Naïve Bayess三网的分类器构建算法。首先,采用三种不同的分类算法训练出不同的弱分类器;然后,通过权重在分配策略,增加患病样本被错分健康样本的权重,减小健康样本被错分的患病样本的权重;最后,通过调整后的权重重组弱分类器,达到构成一种强分类器。利用UCI (University of California, Irvine)数据库中的威斯康星乳腺癌数据进行算法对比验证,实验结果表明:本文所提出分类模型优于单一算法。
In the intelligent diagnosis of breast cancer, the classifier is not stable and the sample distribution adaptability is poor. This paper proposes a classifier construction algorithm based on AdaBoost ensemble BP, RBF and Naïve Bayes. First, three different classification algorithms are used to train different weak classifiers. Then, by means of weight redistribution strategy, the weight of the diseased samples in which are misclassified is increased and reduces the weight of healthy samples misclassified to diseased samples. Finally, a strong classifier is constructed by reorganizing the weak classifier with the adjusted weights. The comparison and verification of the algorithm based on the Wisconsin breast cancer data in UCI database show that the proposed classification model is superior to the single algorithm.
出处
《计算机科学与应用》
2019年第12期2293-2302,共10页
Computer Science and Application