BP改进算法在哮喘症状-证型分类预测中的应用  

Application of BP improved algorithm in asthma symptoms-syndromes classification prediction

作　　者：董国华[1] 陈亚楠[1] 朱习军[1] 

机构地区：[1]青岛科技大学信息科学技术学院,山东青岛266061

出　　处：《计算机工程与设计》2017年第1期215-219,共5页Computer Engineering and Design

基　　金：山东省重点研发计划基金项目(2015GSF119016)

摘　　要：针对BP算法学习率需要人为不断调试且收敛速度慢的缺点,通过对其算法性能进行分析,提出一种基于竞争学习与学习率自适应的改进BP算法,即CAL-BP算法。将改进算法用于哮喘症状-证型的分类预测实验中,将BP算法与CAL-BP算法对哮喘症状-证型数据的训练效果做了对比,实验结果表明,采用改进的CAL-BP算法训练数据时收敛速度更快,识别率更高。To get rid of the disadvantages of BP algorithm that learning rate requires human constantly to debug and the convergence is slow, an improved BP algorithm based on competitive learning and adaptive learning rate was proposed, namely CAL-BP algorithm. The improved algorithm was used in asthma symptoms-syndromes class prediction and compared with two algorithms for asthma symptoms-training effects. Results show that when CAL-BP algorithm used to train data, the convergence is faster and the recognition rate is higher.

关 键 词：BP神经网络 竞争学习 学习率自适应 CAL-BP算法 哮喘症状-证型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]