CAJ | 학술논문

提出了一种基于粗糙集和遗传算法的改进BP神经网络算法.该算法首先对原始数据集进行属性约简,优化BP神经网络的输入变量;然后利用遗传算法全局搜索的特点,优化BP神经网络初始权重和闻值.将改进BP神经网络算法应用于客户分类,训练误差为5.92×10~(-12),测试总误差为0.000 23;而改进前的一个比较理想的训练结果的训练误差为0.001 6,测试总误差为0.073.Matlab仿真表明改进的BP神经网络算法有更好的训练精度和泛化能力.
제출료일충기우조조집화유전산법적개진BP신경망락산법.해산법수선대원시수거집진행속성약간,우화BP신경망락적수입변량;연후이용유전산법전국수색적특점,우화BP신경망락초시권중화문치.장개진BP신경망락산법응용우객호분류,훈련오차위5.92×10~(-12),측시총오차위0.000 23;이개진전적일개비교이상적훈련결과적훈련오차위0.001 6,측시총오차위0.073.Matlab방진표명개진적BP신경망락산법유경호적훈련정도화범화능력.
An improved BP neural network algorithm is proposed based on rough set and genetic algorithm. Firstly, attribute reduction is carried out on original data sets, and the input variables of the BP neural network are optimized. Then, based on the global searching characteristic of genetic algorithm, the original weight and threshold of the BP neural network are optimized. Finally, the improved BP neural network is applied to the custom classification, the training error is 5. 92×10(~-12), and the total test error is 0. 000 23, while an ideal training result's training error before improvement is 0. 001 6, and the total test error is 0. 073. Through the simulation of Matlab, it indicates that the improved BP neural network algorithm has better training precision and extensive ability.