A PCA and neural networks based method for soil fertility evaluation and production forecasting

This paper mainly focuses on the evaluation of the soil fertility levels based on the principal component analysis (PCA) method and production forecasting by neural networks. By combining these two methods (the PCA and the neural networks), we propose a model to describe the relationship between the soil fertility and the crop yield, and present predictions on the yield under different fertilizer models. Some experiments are also given, demonstrating the validity of the combination method. Results show that the proposed model could improve the evaluation accuracy, and optimize the data structure of the neural network model.