Using Explainable Stock Market Prediction Models to Teach Financial Literacy to Students with Disabilities

Abstract

Predicting the stock market is a challenging endeavor in financial learning due to various factors affecting stock prices such as historical trends, technical indicators, macroeconomic conditions, volatility and investor sentiment. In general, predictions studies have focused on the accuracy of predictions, but little focus has been placed on the explainability of prediction models to help students with disabilities and learners in inclusive education settings in their financial literacy. This study proposes an interpretable hybrid ensemble learning model as an educational tool to understand the short-term stock market behaviour. It includes OHLCV data, technical indicators, macroeconomic variables, India VIX, sentiment-based features, missing value treatment, outlier treatment, feature engineering, MinMax scaling, sliding window sequence creation, comparative model training, model evaluation (ensemble), statistical testing, and explainability analysis. This data set has 10 representative stocks of Nifty-50 from 2014 till 2024. The reliability of forecasting accuracy for the selected final model was confirmed with the results, as the model was able to perform accurate forecasting of return on all stocks with MAE, RMSE, MAPE and DA. Lagged price and technical variables were the most influential variables found by SHAP and feature-importance analysis. The study illustrates the potential of explainable AI as a tool for financial literacy education in user-friendly and inclusive learning spaces.