Enhancing Investment Decisions: A Machine Learning Approach to Recommending Stocks and Mutual Funds

Financial literacy is an essential competency in contemporary society, enabling individuals to make informed financial decisions and attain economic well-being. Nevertheless, despite its importance, attaining financial literacy continues to pose a significant problem for numerous individuals. Multiple factors contribute to this challenge, including the intricacy of financial ideas and the extensive amount of information accessible. The financial landscape is ever-changing, with new products, laws, and investing methods frequently arising. Adapting to these changes can be daunting for the typical person. To tackle this issue, this paper investigates various machine learning and deep learning algorithms and techniques with the goal of developing a recommendation system that can suggest stocks and mutual funds in the Indian financial market to users. Recommendation systems represent a successful commercial application of machine learning. It selectively filters information based on the user’s preferences from the vast array of possibilities and suggests options that are optimal or highly probable for the user. We employed Cosine similarity and Pearson correlation to ascertain similarity metrics between two stocks and mutual funds, in conjunction with a Euclidean distance-based method, KNN. We have incorporated three clustering algorithms: K-Means, Agglomerative, and DBSCAN to categorize related mutual funds. We used LSTM for predicting stock prices and Autoencoders for making recommendations. Among all clustering techniques, DBSCAN had the highest silhouette score of 0.85. The LSTM model exhibited a mean training loss of 0.000226 and a mean validation loss of 0.000658 for Nifty-50 businesses, derived from training and testing performed on data spanning the last 20 years.