Proposing Investments Based on Fundamental Analysis

Recent years have witnessed a surge in interest in utilizing machine learning techniques for stock prediction, with studies showcasing their effectiveness in analyzing historical stock data. This thesis focuses on the Nifty 50 stocks and utilizes 22 years’ worth of quarterly financial data for analysis. Six machine learning algorithms were explored, including XGBoost, Random Forest, LightGBM, CatBoost, MultiVariate Linear Regression, and AdaBoost, with an approach that combines fundamental and technical analysis, prioritizing fundamental analysis. Rather than adopting a global training approach, the researchers opted for local training, developing individual models for each Nifty 50 stock to identify the best-performing model for each specific stock. The analysis demonstrated that machine learning algorithms can establish meaningful connections between past financial data and projected revenues, thus providing valuable economic insights and the potential for stock market predictions based on a fusion of fundamental and technical analysis. Moreover, the thesis extended its research to include the development of a rule-based machine learning model to evaluate the financial health of Nifty 500 companies. The model employed predefined rules and criteria based on established financial principles and industry standards, utilizing various financial indicators such as revenue growth, profitability ratios, liquidity ratios, and debt levels. The model’s output was a health score, enabling objective comparisons and rankings of companies within the Nifty 500 index, empowering investors, analysts, and stakeholders in making informed investment decisions. Further details regarding the specific financial indicators and rules applied in calculating the health scores for Nifty 500 stocks can be found in the thesis.