摘要建立定量构效关系(QSPR)模型探索影响有机化合物热容的结构因素,为热容的预测提供理论依据。从622个有机化合物的样本集中随机选择497种作为训练集用于建立模型,125种作为测试集用于验证评价模型。应用DRAGON 2.1软件计算有机物分子的分子描述符及运用蚁群算法筛选出6个分子描述符作为模型的输入变量,随后使用共轭梯度法和支持向量机法分别建立了线性模型(ACO-CG)和非线性模型(ACO-SVM)。预测结果表明:所建模型稳健,泛化能力强,预测误差小,预测效果令人满意,并且共轭梯度模型(R2=0.9494)和支持向量机模型(R2=0.9501)能够反映出有机物热容和其分子结构的关系,结果具有很高的精度。64160
毕业论文关键词 定量构效关系,蚁群算法,共轭梯度,支持向量机
毕业设计说明书(论文)外文摘要
Title Quantitative Structure-Property Relationship for the prediction of heat capacity of organic compounds
Abstract Quantitative structure-property relationship (QSPR) model was established to explore the factors that influence the structures of heat capacity of a set of organic compounds in order to provide the theoretical foundation for the prediction of heat capacity. The dataset of 622 kinds of the organic compounds was randomly pided into a training set (497) for model establishment and a testing set (125) for model validation. Dragon 2.1 software was applied to calculate molecular descriptors and ACO algorithm was adopted to select molecular descriptors and as a result six molecular descriptors were selected as the input variables of QSPR model. Conjugate gradient method and support vector machine method were utilized to produce linear model (ACO-CG) and nonlinear model (ACO-SVM) respectively. The results showed that the QSPR model was of great robustness, outstanding generalization ability and small prediction error. The predicting results was satisfactory, and conjugate gradient model(R2=0.9494)and support vector machine model(R2=0.9501) could build the relationship between the heat capacity and their structures with high accuracy.
Key word QSPR; heat capacity; ACO algorithm; conjugate gradient; support vector machine
1 引言 1
1.1 课题背景和意义 1
1.3 本论文工作 4
2 研究方法介绍 5
2.1 QSPR的介绍 5
2.2 蚁群算法介绍 7
2.3 共轭梯度法 9
2.4 SVM介绍 11
3 热容的QSPR研究 14
3.1 研究步骤 14
3.2 QSPR研究结果 17
3.3 与已有模型的比较 24
4 结论 26
致 谢 27
参考文献 28
附表 32
1 引言
1.1 课题背景和意义
化学一直被认为是一门实验科学,注重实验方法和实验结果。伴随着量子化学——理论化学的一个分支学科的出现,这一观点才被打破。量子化学从微观角度揭示了稳定和不稳定分子的结构,但是,物质分子的结构与其性质之间有何种联系,一直未能有很好的解释。随着科技的发展,计算机技术逐渐应用到化学学科中来。20世纪60年代以后,量子化学计算方法得到极大的发展,其中出现了从头算方法、半经验算法等方法。20世纪70年代中期,Hansch和Free-Wilson等人借助计算机技术,建立了物质分子结构及其性质关系的表达式,开创了定量构效关系(Quantitative Structure Property Relationship,QSPR)研究的先河。论文网