基于支持向量机的卤代甲烷Δ_fH_m~θ模型的构建

Modeling the Δ_fH_m~θ of halomethanes based on support vector machines

应用G98W程序包中的AM1方法,对43种卤代甲烷分子进行优化计算,所得量化参数作为支持向量机的输入向量与ΔfHθm进行建模,并与反向传播人工神经网络相比较,用留一法交叉校验对实验值进行预测。结果表明,支持向量机所建模型预测值具有良好的泛化能力,可用于卤代甲烷其他理化性质的预测。

The quantificatinonal parameters obtained from optimizing calculation to 43 methane molecular by AM1 method in G98W software packet were used as the inputs of support vector regression（SVR） and the model of △fH^θm was established. It was compared with the atificial neural network with error back - propagation（ANN- BP）, the experiment value was predicted by using of the leave - one - out method.The results indicated that it had good generalizing ability for predicting data, and would be used to predict the other characters of physical chemistry of halomethanes.