Event-based stock market prediction

There are various studies on the behavior of the market. In particular, derivatives such as futures and options have taken a lot of attentions, lately. Predicting these derivatives is not only important for the risk management purposes, but also for price speculative activities. Besides that accurate prediction of the market’s direction can help investors to gain enormous profits with small amount of capital [Tsaih et al., 1998]. Stock market prediction can be viewed as a challenging time-series prediction [Kim, 2003]. There are many factors that are influential on the financial markets, including political events, natural disasters, economic conditions, and so on. Despite the complexity of the movements in market prices, the market behavior is not completely random. Instead, it is governed by a highly nonlinear dynamical system [Blank, 1991]. Forecasting the future prices is carried out based on the technical analysis, which studies the market’s action using past prices and the other market information. Market analysis is in contradiction with the Efficient Market Hypothesis (EMH). EMH was developed in 1970 by economist Eugene Fama [Fama, 1965a, Fama, 1965b] whose theory stated that it is not possible for an investor to outperform the market because all available information is already built into all stock prices. If the EMH was true, it would not be possible to use machine learning techniques for market prediction. Nevertheless, there are many successful technical analysis in the financial world and number of studies appearing in academic literature that are using machine learning techniques for market prediction [Choudhry and Garg, 2008]. One way to forecast the market movement is by analyzing the special events of the market such as earnings announcements. Earnings announcement for each stock is an official public statement of a company’s profitability for a specific time period, typically a quarter or a year. Each company has its specific earnings announcement dates. Stock price of a company is affected by the earnings announcement event. Equity analysts usually predict the earnings per share (EPS) prior to the announcement date. In this project, using machine-learning techniques, we want to predict whether a given stock will be rising in the following day after earnings announcement or not. This will lead to a binary classification problem, which can be tackled based on the huge amount of available public data. This project consists of two major tasks: data collection and application of machine learning algorithms. In §2, we will discuss our efforts to collect the required data. In continuation in §3, features definitions and selections are described. We have considered and discussed different machine learning algorithms in §4. In §5, the summary of our results and possible extensions are explained.