基于人工蜂群算法优化SVM的鱼类毒性预测模型  

Fish toxicity prediction model based on artificial bee colony algorithm optimized SVM

作　　者：鄂旭[1] 周艺 李俏竺 E Xu;ZHOU Yi;LI Qiaozhu(College of Information Science and Technology,Bohai University,Jinzhou 121013,China)

机构地区：[1]渤海大学信息科学与技术学院,辽宁锦州121013

出　　处：《渤海大学学报（自然科学版）》2021年第4期357-362,共6页Journal of Bohai University:Natural Science Edition

基　　金：辽宁省社会科学规划基金项目(No:L19BGL016)。

摘　　要：定量构效关系(QSAR)是预测鱼类急性毒性的常用手段,其中多维构效关系为非线性回归问题,在处理这种非线性问题时往往存在一定的局限性.为了更加准确地预测鱼类急性毒性,本文利用了在非线性情况下表现良好的支持向量机(SVM)进行预测,提出了一种基于人工蜂群算法(ABC)优化支持向量机的鱼类毒性LC50预测模型,并与SVM、GA-SVM、PSO-SVM模型的结果进行对比.结果表明,ABC-SVM模型的预测准确率已达到了85.166%,在迭代多次的情况下,运行时间仅46.518秒,具有最高的预测精度、最小的误差,是相对高效的一种鱼类毒性预测方法.Quantitative structure-activity relationship(QSAR)is a commonly used method to predict the acute toxicity of fish.Among them,the multi-dimensional structure-activity relationship is a nonlinear regression problem.There are often certain limitations when dealing with this nonlinear problem.In order to predict acute toxicity of fish more accurately,a fish toxicity LC50 prediction model based on artificial bee colony algorithm optimized support vector machine is proposed,and its result is compared with that of SVM,GA-SVM and PSOSVM model.The results show that the prediction accuracy rate of ABC-SVM model has reached 85.166%.In the case of multiple iterations,the running time is only 46.518 seconds.And the model precision is the highest and the error is the smallest,which is a relatively effective method for forecasting acute toxicity of fish.

关 键 词：支持向量机 人工蜂群算法 鱼类毒性预测 预测精度 运行效率 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]