ADDRESSING CHALLENGES IN STOCK SELECTION: A FINANCIAL DECISION SUPPORT SYSTEM APPROACH

Abstract

Accurate portfolio selection is a critical aspect of investment management. Traditional stock screeners are commonly used to filter stocks based on key financial metrics. However, the selection process becomes challenging when using price-dependent indicators like the Price to Book ratio and Price to Earnings resulting in either an overwhelming number of options or too few viable choices. To address this challenge, this research paper proposes the design of a Financial Decision Support System (DSS) that combines traditional fundamental analysis with machine learning (ML) to score stock valuations and identify the most undervalued stocks from a large list obtained from stock screeners.  Logistic regression and Random Forest models were employed and results were compared for the analysis. 140 stocks and past 10-year data are given as program inputs. When compared to the Logistic Regression model, the Random Forest model fared better with Accuracy, F1-Score, Recall and Precision values of 84.9%, 0.902, 90.0% and 90.4%. The Random forest model when fed with Out of Time Data for the selected list of 140 stocks, 57 stocks scored above 80% and 15 scored above 90% as highly undervalued.

Keywords – Machine Learning (ML); Decision Support System (DSS); Equity Portfolio selection; Stock Intrinsic  value; Stock Valuation scoring; Financial Engineering.

JEL Classification codes: G11; G17; G29