Improving prediction efficiency of Chinese stock index futures intraday price by VIX-Lasso-GRU Model

Abstract

With T+0 and short selling mechanism, the stock index futures are attractive to short-term traders in China, where stocks cannot be liquidated within the day and are difficult to short. So in terms of futures, how to improve the accuracy and speed of intraday price forecasting always fascinates short-term traders and researchers. Here we propose a novel forecasting model, VIX-Lasso-GRU Model, which based on the gated recurrent unit (GRU) by adding VIX information and a method called Least absolute shrinkage and selection operator (Lasso). The volatility index (VIX) can reduce the prediction errors and the Lasso algorithm significantly improve the training speed of the model. We predict the 5-minute closing prices of three datasets of index futures by VIX-Lasso-GRU Model. Comparing to the pure GRU and LSTM, we find that this new prediction model can improve the prediction efficiency with faster speed and higher accuracy.