Association Rule Mining for Portfolio Management A Machine Learning Perspective

This paper explores the application of machine learning techniques, specifically association rule mining, for portfolio management in the Indian stock market. Using data from 149 companies across eleven key sectors of the National Stock Exchange (NSE), we implement APRIORI and CHARM algorithms to discover frequent itemsets and association rules that reveal significant cross-sector and intra-sector relationships. These insights help identify correlated stocks and sectors, allowing for the construction of diversified investment portfolios with higher confidence levels. The analysis includes experimental determination of minimum support and confidence thresholds to prune uninteresting rules and generate meaningful patterns. The portfolio created on this basis is evaluated over a one-year holding period, with results showing substantial returns compared to initial investment. This study demonstrates the efficacy of leveraging machine learning-driven association rule mining on big data to enhance stock portfolio formation and performance analysis, providing a data-driven approach to financial decision-making in dynamic markets