Research on the precise fertilization based on MapReduce model for BP neural network field

With the help of MapReduce powerful parallel computing ability and good extensibility, we try to solve the bottleneck problem of traditional BP neural network in dealing with the big data for the training sets in this paper. Through the experimental data for the farmland fertilizer effect, we propose that fertilizer rate is taken as input for the neural network, and ultimately yield is taken as the output, then building the precise fertilization model. By solving the nonlinear programming problem, the model can get the maximum yield and the optimum fertilizer rate at the same time, and can solve the estimating crop yield problem. In terms of predicting accuracy, the fertilization model result basing on the big data training sets is much better than the small data sets result obviously, and our proposed model can effectively guide the precise fertilization.