Precision Balancing: Machine Learning Models and Selective Strategies for Intraday Trading Success

This paper presents a two-step analysis to maximize gains through predictive methodologies for the stock prices of the top 10 contributors to the NIFTY 50 index. Initially, we conduct a comprehensive evaluation of six machine learning models—K-Nearest Neighbors (KNN), Random Forest Regressor (RF), Decision Tree (DT) Regressor, Support Vector Regression (SVR) without initialization, SVR (y=x) with initialization, and Linear Regression (LR)— within the context of intraday trading. LR and SVR (y=x) demonstrate superior performance compared to other models. Subsequently, we utilize findings from the LR model to formulate and assess five distinct intraday trading strategies (HIGH_SELL, LOW_BUY, ModHIGH, ModLOW, and HYBRID). Emphasis is placed on maintaining a delicate balance between capturing favorable trades and exercising selectivity. The HYBRID strategy proves effective, achieving a 140% gain over 30 trading days. ICICIBANK shows the maximum gain at 254.53%, LT follows this with a gain of 216.31%. The evaluation in the second step relies on a novel metric, percentageGainMin, which addresses the practical scenario of reusing capital for intraday trading. This research contributes insights for both researchers and practitioners, offering a comprehensive exploration of machine learning models and effective intraday trading strategies in the dynamic landscape of financial markets.