Multiple kernel support vector regression for economic forecasting

Economic forecasting has become an important research topic in field of management science. Economic operation is a complex and changeable thing. There are many factors, which impact development of economy positively or negatively. This fact makes the economic system have dynamic, non-linear and uncertain characteristics. In this paper, a forecasting method is proposed for economic research, based on multiple kernel support vector regression. In the proposed method, we provide the forecasting framework for economy by means of multiple kernel support vector regression and multiple kernel learning mechanism. To validate the effectiveness of the proposed method, experiments are conducted on total production amount data from Chinese first and second industry. The numerical result shows that the proposed method greatly outperform conventional BP neural network and support vector machine with simple kernel in terms of forecasting performance.