四川农业大学学报 ›› 2006, Vol. 24 ›› Issue (01): 55-60.doi: 10.16036/j.issn.1000-2650.2006.01.013

• 研究论文 • 上一篇    下一篇

多环芳烃致癌性预测模型比较研究

印家健1, 邹平1, 祁正兴2, 王显祥1   

  1. 1. 四川农业大学 生命科学与理学院, 四川 雅安 625014;
    2. 青海民族学院 化学系, 青海 西宁 810007
  • 收稿日期:2005-10-13 出版日期:2006-03-31 发布日期:2017-03-04

Comparison Study of Predicting Model of Polycyclic Aromatic Hydrocarbons Carcinogenic Properties

YIN Jia-jian1, ZOU Ping1, QI Zheng-xing2, WANG Xian-xiang1   

  1. 1. College of Biology and Science, Sichuan Agricultural University, Yaan 625014, Sichuan, China;
    2. Department of Chemistry, Qinghai Nationality University, Xining 810007, Qinghai, China
  • Received:2005-10-13 Online:2006-03-31 Published:2017-03-04

摘要: 基于量化参数和拓扑指数,分别采用主成分分析和相关分析进行变量筛选,运用留一交叉检验法,引入模型预测性能的评价体系和指标,比较了支持向量机(SVM)、Fisher判别法和K-最近邻法等方法构建的多环芳烃致癌性二值分类预测模型,结果显示SVM要好于其他方法,说明SVM算法具有较强的稳健性和良好的泛化能力,能够用于多环芳烃致癌性的二分类和预测。

关键词: 多环芳烃, 致癌性, 预测, 支持向量机

Abstract: Based on quantum-chemical parameter and molecular topology index, variable is selected by principal component analysis(PCA) and correlation analysis. Leave-one-out and cross calibration method is adopted and assessment system and index of model prediction performance are introduced, modeling and comparing sixty-seven polycyclic aromatic hydrocarbons carcinogenic properties in Support vector classification, Fisher and KNN method are compared. The experiment indicates supporting vector machine possesses better robusticity and generalization capability. It is used in classification and prediction of polycyclic aromatic hydrocarbons carcinogenic properties.

Key words: polycyclic aromatic hydrocarbons, carcinogenicity, forecast, support vector machine

中图分类号: 

  • O64