Non-Stationary Forecasting
Time series forecasting plays a critical role in understanding and predicting dynamic systems across various domains, including healthcare, and environmental monitoring. However, accurately forecasting non-stationary signals—data characterized by changing statistical properties over time—remains a significant challenge. This project aims to investigate the effectiveness of current machine learning-based forecasting models in predicting non-stationary events. By evaluating the performance of established methods such as GPT-based approaches on real-world non-stationary datasets, this study seeks to identify their strengths, limitations, and applicability.
