【作者】 苏俊杰； 钟秋海； 许继平；

【Author】 SU JunJie 1,2,ZHONG Qiuhai 1,XU Jiping 3 1.College of Automatic Control,Beijing institute of technology,Beijing 100081,P.R.China2.Wuzhou Police Fire Brigade,Wuzhou,Guangxi 543000,P.R.China3.College of Information Engineering,Beijing Technology and Business University,Beijing 100037,P.R.China

【机构】 北京理工大学自动化学院； 梧州公安消防支队； 北京工商大学信息工程学院；

【摘要】 结合BP网络、径向基函数网络及遗传算法的各自优点,提出一种混合式预测方法——基于自适应遗传算法的径向基函数网络。该算法采用径向基函数网络的学习算法,引入自适应机制的浮点数编码的遗传算法,并将其与梯度下降法混合交互运算。它可以用模型的预测性能作为优化目标,克服了径向基函数网络在现有学习算法上的缺陷。同时,也解决了单独利用遗传算法往往只能在短时间内寻找到接近全局最优解的这一问题。仿真结果表明该算法有较好的预测精度。更多还原

【Abstract】 A hybrid algorithm——RFN network of self-adaptive genetic algorithm was introduced,which combined the excellences of BP network,RFN network and genetic algorithm.The hybrid algorithm adopts the learning rule of RFN network and combines self-adaptive genetic algorithm and gradient descent method.The capability of prediction can be optimized using the hybrid algorithm and the shortcoming of the learning rule of RFN network was overcomed.At the same time,the problem that Global Optimal Solution always cann’t be found only with genetic algorithm was solved.Simulation results show that hybrid algorithm can obtain better forecasting precision.更多还原