Forecasting stock price movement in index using world news

Abstract

Machine Learning and Natural Language Processing have become an integral part of the Financial Markets. These algorithms are being developed every day to utilize the data available with us and provide a prediction regarding the market’s behavior in future. It has become integral for portfolio managers to use these advanced technologies to manage their portfolio to mitigate risk as well as generate superior returns for the investors. This project revolves around the concept of the Efficient Market Hypothesis and uses Sentiment Analysis of world news to predict market movements. Our idea is to develop a model that uses Natural Language Processing techniques and Neural Networks to forecast price movement in the Index. We plan to obtain a functional relationship between top 20 headlines published during the day & predict the price movements of a composite stock index and thus use the data to make predictions regarding future market movements. We propose to use this model to first verify the efficient market hypothesis across US markets. Further, if the data suggests that the financial markets doesn’t follow efficient market hypothesis, we would also like to explore the possibility of creating a trading strategy around this model and verify the feasibility of this strategy on the basis of risk associated with strategy such as variance of the return or maximum drawdown of the returns. Through these measures we wish to quantify whether this trading strategy could actually result in generating alpha for the investor.