A Combination Prediction Model Based on SVM and Its Application in Grain Output

Grain output forecast plays an increasingly important role in national economy. As grain output forecast is a complex nonlinear system and the historical data is typical small sample, the current methods cannot meet the accuracy requirement. To get better prediction results, a new combination model based on support vector machine (SVM) and BP neural network is proposed. In this model, firstly we use SVM to classify the initial sample data. Secondly, the processed data is used in neural network in selecting the model parameters. By this way the learning speeding can be improved greatly. During the forecast operation, the core modules of the two models are organized together organically and the control information can be passed to each other to adjust model parameters. The organization and working pattern of the combination model enhances the stability and accuracy of forecast results.