A quantitative structure-property relationship (QSPR) model was proposed for predicting electric spark sensitivity of 39 nitro arenes. The genetic function approximation (GFA) was employed to select the descriptors that have significant contribution to electric spark sensitivity from various descriptors and for fitting the relationship existed between the selected 8 descriptors and electric spark sensitivity. The correlation coefficients (R2) together with correlation coefficient of the leave-one-out cross validation (Q2CV) of the model are 0.924 and 0.873, respectively. The model is highly statistically significant, and the robustness as well as internal prediction capability of which is satisfactory. The results show that the predicted electric spark sensitivity values are in good agreement with the experimental data.