Influence on normalization and magnitude normalization for price forecasting of agricultural products

In order to obtain suitable model and increase accuracy in price forecasting, data pretreatment approach via data normalization and the order of magnitude normalization which has the effect on price forecasting is proposed in this paper. The new data preprocessing method that normalized or normalizing the order of magnitude of the original data based on the proposed approaches are described in detail. The data of agricultural products extracted from Website based on improved Back Propagation (BP) neural network and Support Vector Machine (SVM) are utilized the proposed normalization methods and obtain the different results. Experiments demonstrate that the best forecasting average accuracy based on SVM is improved 0.33 percent by normalization and 0.35 percent by normalization of magnitude 10 and the best forecasting average accuracy based on improved BP neural network with no normalization is the best one, but the best of normalization of magnitude 100 cab be lifted 0.66 percent compare with the average accuracy of no normalization. Experiments demonstrate that the proposed approach performance and proves a new data pretreatment method via normalization is meaningful and useful for the model research of price forecasting. © 2013 Asian Network tor Scientitic Information.