Interpretability in Financial Forecasting: The Role of eXplainable AI in Stock Market

The financial services sector, particularly asset man-agement companies, is subject to stringent regulations, with investment decisions undergoing continuous scrutiny by com-mittees. While Artificial Intelligence (AI) has the potential to revolutionize financial services, its adoption in complex market analysis is hindered by its black-box nature. This paper addresses the challenge of time series forecasting for predicting market stability, leveraging comprehensive data from the Pakistan Stock Exchange. We conducted an experimental comparative study to classify time series data and to enhance model interpretability. We proposed an Explainable AI (XAI) framework using post-hoc techniques such as Local Interpretable Model-agnostic Ex-planations (LIME) and SHapley Additive exPlanations (SHAP). Our ConvlD-BiLSTM architecture achieved an accuracy of 83%. LIME interpretations revealed that moving averages were the most significant contributors to the predictions. The proposed framework provides key feature contributions, making model decisions transparent and interpretable for stock market partie-inants.