Effect of Magnitude Differences in the Raw Data on Price Forecasting Using RBF Neural Network

In order to improve the stability on price forecasting with the RBF neural network, the researching and improvements on the raw data processing method for price forecasting using neural network are discussed in this paper. The magnitudes of the raw data are normalized to a relatively small magnitude, in other words, reducing the magnitude of the raw data to reduce the fluctuation range of the neural network training data. According to this idea, the different experiments are implemented with RBF neural network, which is using weekly price of 10 different agricultural products by enhancing and reducing the magnitude of the raw data. Comparing the experimental results with the actual data of three different experimental approaches, the experiment with the data reduced gets the best forecast results. And the forecasting average accuracy obtains 96.8 percent. Using the proposed improvements method not only improves the stability of the price forecasting by RBF neural network, but also being able to meet the needs of practical application.