SVR Model Used for Economic Fluctuation Analysis

Abstract

The purpose of this study is to find an optimal algorithm for the prediction of market value and the analysis of economic fluctuations. We propose an ensemble learning algorithm based on SVR and apply it to market value prediction and economic fluctuation analysis. It was found that in most situations, the smaller the window value of short-term learning model is, the smaller the weight of long-term learning model is, and the better the performance of ensemble learning model is. However, with the decrease of weight value, ensemble learning model will have the problem of over-fitting, which makes the performance of the model decline. This paper proposes a market value forecasting model based on long-term and short-term ensemble learning. In the theory of SVR model, the validity and superiority of the model are verified through a large number of experiments. [1]