Stock Market Analysis and Prediction

Stock market analysis is a widely studied problem as it offers practical applications for signal processing and predictive methods and a tangible financial reward. Creating a system that yields consistent returns is extremely challenging and is currently an open problem as stock market prices are extremely volatile and vary widely both within a given stock and comparatively amongst many stocks. Further, stock market data is influenced by a large number of factors including foreign and domestic economies, trade agreements, wars, seasons, and even day of the week [12, 3]. Any trading strategy must of necessity balance desire for immediate returns with the possibility of larger payoffs in the future, and many different approaches towards prediction have been attempted including neural networks and fuzzy reasoning [1], support vector machines [9], and even attempting prediction using data-mining techniques over textual data in financial news [14]. In this paper, we examine the abilities of linear regression, random forests, and support vector machines (SVM) with SMO to predict future prices and trends in a variety of stocks.