Quantitative structure activity relationships (QSAR) studies were performed on the radical scavenging activities of a set of compounds consisting of various types of antioxidant families. The predicting five parameter models correlating selected descriptors, derived from the 2D and 3D representations of molecules and antioxidant activity, were set up using multiple linear regressions (MLR) and a multilayer perceptron neural network (MLP-NN), separately. The best obtained model had statistics of R-2 = 0.968 and q(2) = 0.898 for the MLP-NN model and R-2 = 0.902 and q(2) = 0.862 for the MLR model. The chemical applicability domains of these models were determined via a leverage approach. The obtained result indicated that the proposed models can be successfully used for predictions of radical scavenging activities of new antioxidants.