自适应遗传算法优化支持向量回归的水质预测  

Water quality prediction based on adaptive mechanism for GA-SVR

作　　者：吕丹 龙华[1] 刘洋[1] 赵继东 杜庆治[1] LV Dan1,LONG Hua1,LIU Yang1,ZHAO Ji-dong2,DU Qing-zhi1(1. School of Information Engineering and Automation, K u nm in g U n iv e rs i ty o f Science a n d T e c h n o lo g y, K u nm in g 650 50 4 , C h in a ; 2. The Protect ion Agency of Erhai Lako in Dali Prefecture, Dal i 6 7 1 0 0 0 ,Y u n n a n P ro v in c e, C h in a)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650504 [2]大理州洱海流域保护局,云南大理671000

出　　处：《信息技术》2018年第4期37-40,共4页Information Technology

基　　金：国家自然科学基金资助项目(61070161;61070158);2014云南省科技厅基金(2014RA051)

摘　　要：文中提出了一种动态改进的遗传算法和支持向量回归机相耦合的水质预测方法,改进了传统遗传算法中交叉和变异概率固定的问题,尽可能避免陷入局部最优的问题。在对大理弥苴河水质进行大量实际监测的基础上,分别采用BP神经网络,遗传算法优化的支持向量回归机和自适应遗传算法优化的支持向量回归机3种模型的方法,建立了弥苴河水质高锰酸盐含量的的预测模型。通过数据预处理,筛选了60天的数据进行训练学习和测试。通过对三个模型的预测误差分析对比,可以得出自适应遗传算法优化支持向量回归的预测模型精度更高。This paper proposed a method of water quality prediction that combined adaptive mechanism for genetic algorithm with support vector regression（ GA-SVR）,which improved the problem of the constant probability rates for crossover and mutation, in order to avoid result from local optimum scenario. Three predition models for CODMn in Miju River were established respectively by using back propagation（ BP） neural network,GA-SVR and adaptive mechanism for GA-SVR on the basis of monitoring. 60 days of data were selected for learning and testing. It approved that the model of water quality prediction based on adaptive mechanism for GA-SVR can provide better accuracy.

关 键 词：水质预测 支持向量回归机 自适应遗传算法 预测模型 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]