Time series forecasting plays a critical role in both statistical learning(SL) and machine learning (ML), with applications spanning industries such as supply chain management and finance. Accurate long-term forecasting is essential for optimizing resource allocation, improvingndecision-making, and addressing dynamic challenges arising from evolving data trends and seasonal variations.

Recent advancements in time series forecasting, particularly through Automated Machine Learning (AutoML), have demonstrated significant effectiveness in both short-term and longterm predictions for univariate and bivariate time series. This effectiveness is even more pronounced in multivariate time series forecasting. This presentation will explore these advancements, highlighting the superior accuracy of AutoML approaches over traditional statistical methods such as Autoregressive Integrated Moving Average (ARIMA) and Vector Autoregression (VAR), especially in scenarios where seasonality and trends have shifted over time.

Furthermore, the proposed approach leverages machine learning to automate the entire forecasting process, enabling researchers and data scientists—regardless of their expertise in time series analysis—to effectively apply these methods with minimal manual intervention.