Chandan Maka

There are lots of stocks that are listed in various stock exchanges and that number can reach in thousands. For most of the investors, investment firms and analysts the main challenge is to find out the stocks that can be classified as fundamentally strong companies within the market. With an attempt to find the solution to the above problem, previously multiple works have been done to classify the stocks or select the stocks of a companies whose financial status is strong. This practicum project seeks to classify the stocks with machine learning models and plenty of the fundamental metrics associated with those stocks collectively. This approach utilizes the fundamental financial metrics related to profitability, liquidity, leverage, cash flow quality, efficiency, capital investment, cost structure and shareholder-related measures derived from publicly disclosed financial reports. The goal is to find out the stock’s financial status with the help of machine learning model that has been trained under supervision of collected and prepared data. This practical approach will find its potential in the financial domain of equity, portfolio and risk.

For any investors, making an accurate trading decision is highly complicated task because of the high volatility nature of financial markets. Traditional machine learning models rely on historical data and statistical methods and struggle to adapt to changing market environment and conditions. This project aims to develop a reinforcement learning (RL) approach for stock trading that utilizes various technical indicators to improve trading decisions and performance. The primary objective of this project is to develop a reinforcement learning (RL) model that can learn optimal buy and sell decisions interacting with the historical prices of stock. The approach focuses on integrating multiple technical indicators such as moving averages, momentum indicators, and volatility measures as input features for the RL model. It will involve data collection, preprocessing, normalization and feature engineering. Lastly, it will also include model training using reinforcement learning algorithms such as Deep Q-Network (DQN)and Proximal Policy Optimization (PPO). Based on performance metrics such as equity growth overtime, the trained models will be evaluated on test data. This project aims to demonstrate how reinforcement learning techniques can be effectively utilized with technical indicators in trading problems and enhance the learning process as well. This project contributes to the development of adaptive trading techniques using RL models in financial markets. This project focuses on the practical implications of reinforcement learning in algorithmic trading.