Financial volatility forecasting based on inter- company connections and support vector machine

Within stock exchanges, trading volume and asset price are considered to be vastly volatile, whereas the effective prediction towards them provides constructive advice for financial practitioners. Meanwhile, due to the fast encroachment of information technology into every level of life, financial news holds its very importance in affecting as well as evaluating the market behaviour. It has been proved that the news volume exhibits an observable impact onto trading volume volatility. On the other hand, nowadays almost no company is isolated from the industry community. Relationship network within stock exchanges, namely that listed companies are considerably inter-dependent and connected, leads to the fact the change of the market behaviour related to one company might impose noticeable changes onto that of another. In this paper, we present an approach to mine the associations between the volatility and the online financial information volume for US market, with the employment of inter-company connections. GARCH theory is adopted in conjunction with support vector machine to achieve a modified non-linear learning model. Empirical studies demonstrate that an adoption of related companies produces a prominent enhancement in forecasting performance towards financial volatility.