Mining associations between trading volume volatilities and financial information volumes based on GARCH model and neural networks

There has been an increasing attention on the influences online financial information has on the financial markets. In the meanwhile, the volatility of trading volumes, just as the volatility of stock returns, has an inseparable association with financial risks. It has been considered that there might exist some direct or indirect correlations between online financial information volumes and financial volatilities, though corresponding quantitative analyses or empirical studies are still absent. In this paper, we introduce a mathematical model utilizing artificial neural networks (ANNs) and GARCH (Bollerslev, 1986) model, in order to mine the associations in between. The rudimentary mathematical basis is the GARCH model, while we introduce the volume of financial information from the Internet as an exogenous input, in conjunction with artificial neural networks as the prediction tool. Since combining ANN and GARCH to probe into the correlations between the aforementioned two is somewhat left untouched, it’s worth mentioning that not only have we realized the prediction of the trading volume volatilities to an acceptable extent; we also have quantitatively analyzed the model’s forecasting ability for the volatility trends. Besides, we further substantiate the impact online financial information has on financial trading volume volatilities via a series of disturbance experiments. Furthermore, we have presented a basic forecasting measure relying on the volatility-clustering feature, and proved that our model significantly outplays this measure in forecasting volatility trend.