In this paper, the implement steps of natural forest protection program grading (NFPPG) with neural network (NN) were summarized, and the concepts of program illustration, patch sign unification and regress, and inclining factor were set forth. Employing Arc/Info GIS, the tree species diversity and rarity, disturbance degree, protection of channel system, and classification management in Moershan National Forest Park were described, and, used as the input factors of NN, the relationships between NFPPG and above factors were analyzed. Through artificially determining training samples, the NFFPG of Moershan National Forest Park was built. Tested with all patches in the park, the generalization of NFFPG was satisfied. NFPPG took both the classification management and the protection of forest community types into account, as well as the ecological environments. The excitation function of NFPPG was not seriously saturated, indicating the leading effect of inclining factor on the network optimization.