Abstract: In order to study the antifungal activity of 2-heteroaryl-4-chromanone derivatives to rice blight and develop the novel bactericide with high activity, the molecular structure of 2-heteroaryl-4-chromanone derivatives were characterized by molecular connectivity index and molecular electrical distance vector. The four-parameter(X₁, M₃₆, M₁₄, M₃₂) QSAR model of pIC₅₀ was constructed by leaps-and-bounds regression(LBR), the traditional correlation coefficient (R²) and the cross-validation correlation coefficient (R^2_\rmCV) of leave-one-out (LOO) were 0.854 and 0.788, respectively. The model showed predictability and robustness by the verification. The four structural parameters were used as the input neurons of artificial neural network, and a 4∶3∶1 network architecture was employed. The total correlation coefficient was 0.983. The results show that there is good nonlinear relationship between the activity (pIC₅₀) and the four molecular structure parameters. According to the results obtained from the structural modifications, four compounds with high antifungal activity were proposed, and it is expected to be confirmed by using biologic relationships.