Interpreting Carbon Price Predictions: A Temporal Analysis with Explainable AI for Market Shock and Recovery

The global disruptions caused by COVID-19 in 2020 profoundly affected various economic and environmental systems, including carbon pricing mechanisms. This study investigates the temporal dynamics of carbon price predictions during March 2020, a period of significant uncertainty, and March 2023, a post-COVID stabilization phase. Using TimeSHAP, we analyze feature-level, event-level, and cell-level explanations to understand how key factors influenced predictions during these contrasting periods. The comparative analysis highlights shifts in feature importance over time, offering insights into the evolving drivers of carbon pricing. This work underscores the importance of explainability in understanding model predictions during periods of volatility and recovery, providing actionable insights for stakeholders in the carbon trading market.