中文摘要：

基于遗传算法（genetic algorithm，GA）的变量筛选和支持向量机（support vector machine，SVM），提出了一种改进的定量结构一性质相关（quantitative structure detonation relationship，QSPR）建模方法——遗传-支持向量机（GA—SVM），并用其建立含能材料的定量结构-爆轰性能关系（QSDR）模型，此外还应用标准SVM方法建立了QSDR模型，并用这2种模型进行呋咱系含能化合物密度的预测，随机选取85％化合物作为训练集，用来建立模型，其余化合物作为测试集来测试模型的预测能力。预测结果的交互检验的相关系数平方分别为0．9887和0．9885，平均相对误差分别为1．16％和2．12％，表明了2种建模方法的有效性。通过对2种模型的预测能力进行比较，GA-SVM方法建立的QSDR模型能更好地预测呋咱系含能化合物的密度，更利于实际应用。

英文摘要：

A modified method to develop quantitative structure property relationship （QSPR） models was proposed based on genetic al- gorithm （GA） and support vector machine （SVM） （GA-SVM）. GA was used to perform the variable selection, and SVM was used to construct QSPR model. GA-SVM was applied to develop the quantitative structure detonation relationship model for energetic materials. The standard SVM was also utilized to construct QSDR prediction models, 85% of the whole date set were taken as training set to construct the model, the rest compounds were used to test the prediction ability of the model. The cross-validation correlation coefficient Rcv^2 is 0. 988 7 and 0. 988 5, and the mean relative error is 1.16% and 2. 12% , for the prediction results respectively. It demonstrates the validity of these two methods. By comparison the stability with prediction ability of the models, it was found that GA-SVM is the optimal method for developing QSDR model for predicting the density of furazan compounds.