Abstract: The crops productivity at the approximate management level reflects the cultivated land fertility grade. On the basis of AHP-fuzzy evaluation method, the ANN-productivity quantitative evaluation model was attempted to establish, and Yutai County, Shandong Provice is taken as case. In this evaluation model, used the relative membership grade to describe the evaluation index, took the crops productivity as the goal output standard, obtained the evaluation model after the neural network training. Compared with the traditional method, the model not only can reflect the non-linear characteristic of the cultivated land fertility, but also the process does not need to determine the weight, and this eliminated human factor influence. More importantly, the objectivity of evaluation results is further increased by using the crops productivity as the goal output. Through contrast with the traditional method, results of this model are consistent with the present evaluation method, and this method has explored a new way for the cultivated land fertility quantitative evaluation.