Prediction model for soil moisture in tobacco fields based on PCA and RBF neural network

In order to build prediction model of the soil moisture so as to easy to plan and manage tobacco planting for tobacco fields, authors presented a method with the principal component analysis (PCA) and radial basis function (RBF) neural network model for predicting the soil moisture of tobacco fields. Firstly, the PCA was used to eliminate the correlation of the initial input layer data so that the problem of efficiency caused by too many input parameters and by too large network scale in neural network modeling could be solved. And then, the prediction model of soil moisture was built through taking the results of PCA as inputs of the RBF neural network. The research result showed that the model proposed had a better prediction accuracy that the average prediction accuracy reached 96.02%, and enhanced 5.20% and 6.06% compared with the conventional back propagation (BP) network and RBF network respectively, which met the requirements of actual tobacco-growing area planting planning and provided a theoretical reference for other types of soil moisture forecasting.