遗传神经网络特征提取及在分类中的应用  

Feature Selection Based on Genetic Neural Network and Its Application in Classification

作　　者：李汪根[1] 黄尧颖[1] 叶小娇[1] 

机构地区：[1]安徽师范大学数学计算机科学学院,芜湖241003

出　　处：《系统仿真学报》2011年第10期2094-2097,共4页Journal of System Simulation

基　　金：国家自然科学基金(61070060);安徽省高校自然科学研究重点项目(KJ2010A140)

摘　　要：将遗传算法和神经网络结合应用于乳腺癌细胞分类,首先利用遗传算法随机提取训练集的属性特征,然后用提取特征后的训练集训练神经网络,最后得到必要的特征子集优化网络结构。仿真实验结果表明,遗传神经网络不仅可以优化神经网络的权值和阈值,还能有效地找出线性可分离特征子集,从而达到降低数据维数并提高分类精度的目的。A genetic algorithm combined neural network with the application in breast cancer cells classification was proposed. The genetic algorithm was used randomly to extract features from training sets. The training sets with extracted features were used to train the neural network. The necessary attributes subsets were used to optimize network structure. The simulation results show that the proposed model can not only optimize the weights and thresholds of neural network, but also effectively find linear separable feature subsets, so as to reduce data dimensions and improve the accuracy of classification.

关 键 词：遗传算法 神经网络 特征提取 分类 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]