| Abstract [eng] |
The intention of this study is to examine the relationship between financial ratios and stock performance in the U.S. stock markets between the years of 2014 and 2018. It reviews the existing literature on financial ratios and their application to financial results prediction, as well as machine learning in the field of stock market prediction. Besides that, the aim is to develop and apply machine learning model for share price prediction that would incorporate a variety of financial ratios as parameters. Master thesis is divided into three sections: literature review, research and findings, and conclusions and suggestions. This paper is centered on the principles of the relationship between financial ratios and stock market returns. Profitability ratios acting as indicators of company success and returns, and mathematical and machine learning-based prediction methods. A review of the literature seeks to assess which financial ratios have a strong correlation with stock returns, regardless of whether mathematical or machine learning approaches are used. Additionally, prior research data and methodologies are explored. After the literature analysis, author goes to inspect the relationship between financial ratios and stock returns in the U.S. stock exchanges from 2014 to 2018 using novel machine learning technique and a large dataset. Quantitative techniques were employed to determine which financial ratios have a strong correlation with stock return. Additionally, a machine learning model was used to investigate which ratios could forecast stock returns, utilizing a set of financial ratios that have been shown in the literature to have a strong correlation with company share price results. The empirical part of this master thesis could be summarized in a few clear steps. The very first phase was data collection, which included obtaining raw data from a third-party service. Followed by workspace creation in Python and developing the dataset. All the empirical portion of the study was done using Python. Next step was exploratory data analysis in order to understand it and prepare for cleaning procedures. After a clear picture of dataset, cleaning was done, where missing values and outliers were handled accordingly. Afterwards, visualizations of relationships and distributions, feature engineering and selection was performed. Finally, prediction modelling using regression and classification methods was performed by employing machine learning algorithm XGBoost. The empirical findings chapter is split into three subsections: overall analysis findings, linear regression and correlation findings, and classification modeling findings using the XGBoost classifier. Finally, author concluded that the top 5 most important financial ratios to predict the following year’s stock price were ROE, PB Ratio, Price Earnings-To-Growth Ratio, Net Profit Margin, and EPS. Moreover, in summary of the all-empirical findings, it can be said that studied financial ratios do possess a relationship with the stock market performance, however not a strong one. Results showed that standalone financial ratios will not provide sufficient material to forecast movement of the stock price. |
