改进的BP神经网络在医疗中的应用研究  

Application of Improved BP Neural Network in Medical Care

作　　者：刘清云[1] 吕艳辉[1] 

机构地区：[1]沈阳理工大学信息科学与工程学院,辽宁沈阳110159

出　　处：《沈阳理工大学学报》2013年第3期48-51,共4页Journal of Shenyang Ligong University

摘　　要：鉴于传统BP神经网络存在诸如收敛速度缓慢和易陷入局部极小值的缺陷,提出在BP神经网络中引入动量因子、陡度因子、自适应学习速率和改进粒子群算法优化BP权值等对传统BP进行改进,使其达到预测误差小、精度高、训练快的目的。改进后用来预测区域内各个年龄段患糖尿病人数百分比。由仿真实验结果可知,改进后的BP神经网络比传统BP神经网络收敛更快、误差更小。In view of the defect of traditional BP neural network, such as slow convergence and easy falling into local mininum value, the traditional BP is improved by introducing mo- mentum factor, steepness factor, adaptive learning rate and optimized BP weights. The pur- pose is to achieve high accuracy. The improved BP is expected to predict the percentage of diabetes carriers in the region. The simulation results show that the improved BP neural net- work has faster converge and smaller margin of error than the traditional one.

关 键 词：BP神经网络 BP算法 粒子群算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]