Stock Returns Prediction Based on Implied Volatility Spread Under Network Perspective

Abstract

Using 50 ETF options data from the Shanghai Stock Exchange as samples, this paper explores the predictive power of option implied volatility spread (IVS) on stock market returns, mainly from a network perspective. In this paper, we first construct a multi-scale data series by wavelet decomposition of the data, and then build a corresponding dynamic complex network on this basis. We analyze the topological features of the network to reveal the dynamic relationship between variables. At the same time, the topological features are used as input variables for machine learning to quantitatively explore the return information contained in the IVS. The conclusions show not only that IVS has the strongest correlation with stock market returns in the medium and long-term, but that the accuracy of IVS prediction is also highest at this time. Furthermore, the GBDT machine learning model is more effective in predicting future stock market returns when using IVS as an indicator.