Abstract—In recent years, global shocks such as the COVID-19 pandemic, geopolitical conflicts, and trade tensions have substantially intensified volatility in financial markets, thereby increasing the frequency of price turning points. This study proposes a novel approach that addresses the growing demand for artificial intelligence models that are not only highly predictive but also interpretable. Traditional econometric models, such as Autoregressive Integrated Moving Average (ARIMA) and Generalized Autoregressive Conditional Heteroskedasticity (GARCH), are effective at capturing linear dependencies but fail to adequately represent nonlinear market dynamics. Meanwhile, machine learning methods, despite their strong predictive power, often exhibit limited interpretability. To bridge this gap, this paper presents an explainable hybrid framework that integrates ARIMA, Long Short-Term Memory (LSTM), and Shapley Additive exPlanations (SHAP). Using daily data from ten major global stock indices from January 1, 2010 to September 30, 2025, the study employs a hybrid labeling approach that combines price extrema, Relative Strength Index (RSI), and Moving Average Convergence Divergence (MACD) indicators. Forecasting accuracy is further enhanced by incorporating ARIMA residual–based feature fusion. The model’s performance is evaluated using both statistical (F1-Score, Precision-Recall Area Under the Curve (PR-AUC), Root Mean Squared Error (RMSE), and economic (Sharpe ratio) metrics within a walk-forward validation framework. The results show that the ARIMA-LSTM-SHAP model achieves an optimal balance between predictive accuracy and economic efficiency, achieving an F1-Score of 0.60 and a Sharpe ratio exceeding 1.5. The findings highlight the framework’s practical relevance for investors and policymakers by facilitating the early detection of turning cycles and supporting risk-informed decision-making.
 
Keywords—turning-point detection, explainable artificial intelligence, behavioral finance, adaptive market hypothesis, economic validation


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).