Associating Financial Trading Volume Volatility and Information Volume based on Neural Network and Support Vector Machine

Within the stock markets, the trading volumes and the asset prices are considered to be highly changeable and unpredictable. However, effective forecasting of the way how they will change guarantees constructive advice for financial practitioners. There are various factors that may have an impact on the movements, one of which is the financial information. On the other hand, financial volatility, be it exhibited by the trading volume or the stock price, is regarded to hold its very importance within stock markets. In this paper, we attempt to employ two approaches to forecast the stock market volatility using the online financial information. Particularly, we delve into the associations between the trading volume volatility and the online financial information volume. GARCH theory is adopted in conjunction with artificial neural network and support vector machine to achieve modified non-linear GARCH-based mining models. Most importantly, financial information downloaded from the Internet is fed into the models as an exogenous input. Relying on these models, we conduct experiments onto the data collected from the US stock markets. As indicated by the comparative study based on the experimental results, while both approaches are eligible enough to obtain acceptable forecasting performances for the volatility, SVM outweighs ANN in this regard, whereas ANN is capable to achieve a more promising forecasting result for the volatility trend. In order to further qualitatively substantiate the impact online financial information has on financial volatility, a series of disturbance experiments are conducted under both the ANN and the SVM settings, whose results show that a noticeable correlation does exist between the aforementioned two.