Predicting stock prices from news data

Abstract

The stock markets are a highly volatile proposition, with the somewhat unpredictable nature of price movements resulting in uncertainties. The biggest of the events, including market crash, economic depressions, etc, to the smallest of events related to politics, economic growth, interest rates, speculations, etc., reflect in price changes for not just a company, but the market as a whole. (Bastianin and Manera 2018). A well-performing stock can face a downturn in its fortunes due to a loss in confidence amongst the investors and shareholders, which could lead to a mass selling off by the shareholders, thus kickstart a domino effect. To a common man and a seasoned investor alike, having some idea about the possibility of upward or downward movements in the stock prices thus assumes high importance. Also, this estimation needs to be made as early as possible since the volume of potential profits in a share transaction can change significantly in a short span of time. Generally, this prediction is made following the trends in the industry, as well as being well versed with the latest happenings and the news. However, physically keeping track of this involves significant time lags, and thus, the potential opportunity could be lost. Thus, a potential area of research translates to building models for accurate predictions of the stock price data using external metrics.