Event driven share price forecasting based on change based impact analysis

Abstract

Investing in stocks is considered one of the riskiest options to invest due to regular unpredictable market fluctuations. It is difficult to forecast stock price variations due to this reason which makes investment or divestment decisions extremely challenging. This paper proposes a mechanism for share price forecasting by quantifying the impact of market externalities such as news events. We propose a novel multivariate approach that forecasts the behavior of stock prices — a projection modified for investor psychology and market features, more reliably compared to existing work. Our mechanism employs a strategy that models stock variations using a physical metaphor employing first-order derivatives of historical stock price and sentiment with respect to time. We do an extended forecast based on the sentimental impact on stock prices in response to an event using Kalman filtering, similarly to a trajectory of a physical object that is subject to a force. The proposed methodology achieves a significant accuracy of up to 97% for two-three days forecasts, which exceeds the forecast accuracy of related work.