基于支持向量机方法的表面活性剂增溶性能分类预测研究  

Prediction of solubilization of surfactants by Support Vector Machine

作　　者：丁振浩[1] 戴幸星[2] 王宇光[2] 史新元[1] 乔延江[1] 

机构地区：[1]北京中医药大学中药学院,北京100102 [2]首都医科大学中医药学院,北京100069

出　　处：《计算机与应用化学》2011年第4期422-424,共3页Computers and Applied Chemistry

基　　金：北京中医药大学中青年教师资助项目"基于计算机模拟的增溶性辅料筛选方法研究"(2009JYB22-JS036);北京中医药大学在读研究生资助项目"基于化学计量学中药增溶性辅料构性关系研究"(2009JYB22-XS032)

摘　　要：为了探讨表面活性剂的增溶性能,计算了表征分子组成和拓扑等特征的148个分子描述符,经属性筛选得到13个描述符,采用支持向量机方法建立了表面活性剂增溶性能分类预测模型。结果表明,该模型预测能力及稳定性良好,5折交叉验证准确率为92.1%,测试集验证准确率为95.1%。用中药皂苷化合物对该模型进行验证,模型验证准确率达93.8%,表明该模型具有良好的推广能力,可为中药增溶性能研究提供指导。In order to predict the solubilization of surfactants,constitutional and topological molecular descriptors,148 in total,were calculated to characterize the structural and physicochemistrical properties.A classification and prediction model was built by Support Vector Machine with 13 molecular descriptors selected by Bestfirst＋CfsSubsetEval method.The model has good stability and predictivity performance.The overall accuracy for training set by means of 5-fold cross-validation is 92.1%.Futhermore,the model was evaluated by using the independent test set.The overall prediction accuracy for the test set is 95.1%.The model was used to predict the saponins for futher research,and the prediction accuracy is 93.8%. The results indicated that the model has good generalization ability and could provide guidance for the solubilization study of traditional Chinese medicine.

关 键 词：增溶 支持向量机 分子描述符 属性筛选 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]