Are Stock Prices Really Random? Exploration Into Support Vector Regression and Stock Market

Abstract

The 2021-2022 supply chain crisis coupled with the current Russia-Ukraine crisis has brought about unique inflationary conditions and market volatility. With U.S. inflation (measured by CPI) rising to 9.1% in July 2022, a 40-year high as of 2022, current market conditions have cultivated distinct trading challenges. This thesis demonstrates that a sub-field of machine learning (ML), Support Vector Regression (SVR), has the capability to forecast future daily stock market prices for investment research. This methodology utilizes technical analysis indicators, like Weighted Moving Average (WMA), Relative Strength Index (RSI), and Average True Range (ATR), and others as predictor variables for the regression model. This thesis intends on performing an SVR on different time horizons of the Standard and Poor's 500 (S&P 500) index equities, utilizing historical data from the time period 1999-2022. Additionally, different kernel functions and hyperparameters will be tested to improve the accuracy of the SVR. In order to assess the effectiveness of the SVR, the model will be contrasted with a Random Walk Model. On a larger scale, this thesis explores the effectiveness of machine learning methodologies to identify profitable trade opportunities through forecasting future market prices from historical data.