In the dynamic and volatile environment of the Indian stock market, identifying accurate buy and sell signals is crucial for investors and traders aiming to maximise their returns. Traditional technical indicators, while useful, often fall short of capturing the multifaceted nature of market movements, particularly during unforeseen events such as the COVID-19 pandemic. This gap in predictive accuracy and reliability highlights the need for more sophisticated analytical methods. The proposed study introduces an innovative approach by integrating extremely randomised trees (ERT), a powerful ensemble learning technique, with conventional technical analysis to forecast Index price movements more accurately. Specifically, the analysis focussed on indices such as Nifty 50 and Bank Nifty during the pandemic period, comparing the performance of ERT against other advanced models including CNN, Bi-GRU, MLF, and AT-GRU-M. The results demonstrated that the Proposed ERT model consistently outperformed other models in terms of precision, recall, F-measure, and overall accuracy, showing remarkable improvements in forecasting capabilities. For instance, in forecasting Nifty 50 and Bank Nifty Indices prices, the ERT model achieved the highest accuracy, surpassing other models by a significant margin. The study’s findings suggest that integrating ERT with technical indicators can substantially enhance the identification of buy and sell signals in the Indian stock market, offering investors a robust tool to navigate market uncertainties and capitalise on investment opportunities more effectively, with accuracy improvements quantified in the range of 5–10% over traditional methods.