International Journal of Scientific Research and Engineering Development

AI-Driven Supply Chain Demand Forecasting uses the Hybrid Time-Series and Gradient Boosting Models which was used to detect the critical challenge in supply chain management due to dynamic market conditions, seasonal variations, and un- certain customer behavior.As traditional time-series forecasting methods fails to capture business factors such as promotions, pricing changes, and sudden demand changes, which leads to inefficiencies to overcome these limitations in this project we use the AI-driven hybrid demand forecasting system that combines time-series models with gradient boosting machine learning techniques. This hybrid model basically divides demand forecasting into two components: base demand estimation and error correction. Timeseries models such as AutoRegressive Integratd Moving Average(ARIMA) [7],Seasonal AutoRegressive Integrated Mov- ing Average (SARIMA) [2], and Long Short-Term Memory (LSTM) [3] are used to capture the real time historical data and captures the trends based on the seasonality, and temporal dependencies.However, deviations between actual demand and time-series predictions occur due to external business factors such as promotions, price changes, and holidays. To change the deviations through this we use the gradient boosting models like eXtreme Gradient Boosting(XGBoost) [4] and Light Gradient Boosting Machine (LightGBM ) [5] which are trained on the residual errors produced by the time-series models.The final demand prediction is obtained by combining the time-series forecast with the adjustment of residual errors. This work facilitates the evolution of resilient supply chains, in addition to the services provided in the emerging deep reinforcement learning models.

Dr.Thalakola Syamsundararao, K.Abhigna, K.Akhila, K. Santhi, A.Charmi Venkata Nagalakshmi,"AI-Driven Supply Chain Demand Forecasting Using Hybrid Time-Series and Gradient Boosting Models" International Journal of Scientific Research and Engineering Development, V9(2): Page(955-959) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.