Precision fertilization is the core content of precision agriculture technology. There was a complex non-linear relationship between crop optimal fertilization and soil fertility. The single model is difficult to accurately describe its complex relationship and change law, making the crop accurate fertilizer is difficult to determine; Neural network technology to solve this problem provides a new way of thinking. But a single radial basis function network (RBF) neural network fertilization model is too dependent on the selection of the hidden layer data center. Therefore, this paper proposes a decision-making technique based on fuzzy C-means (FCM) clustering and RBF neural network fusion algorithm. The fusion algorithm first uses the FCM algorithm to select multiple RBF networks in the training samples. Based on this, the least squares (OLS) training network is used to optimize the data center. Finally, an improved RBF neural network model is established. In this paper, the model is applied to the maize precision operation demonstration base, soil nutrient and maize yield as the input of neural network, using the precision fertilization amount of maize as output. The model of precision fertilization of maize was established. And the model was used to make the precision fertilization decision of maize. Experimental results show: The improved RBF neural network is compared with the traditional BP network to reduce the error by 0.47. Compared with the model, the error of the RBF neural network method is reduced by 0.045. Significantly improve the prediction accuracy, reduce the calculation time. Can effectively guide the precise fertilization.