AI-Driven Demand Forecasting for Multi-Echelon Supply Chains: Enhancing Forecasting Accuracy and Operational Efficiency through Machine Learning and Deep Learning Techniques.

Аннотация

Demand forecasting plays a crucial role in optimizing supply chain operations, particularly in multi- The American Journal of Management and Economics Innovations 74 https://www.theamericanjournals.com/index.php/tajmei echelon supply chains where goods move through Recent advancements in machine learning and artificial various stages, including manufacturers, wholesalers, intelligence (AI) have provided new opportunities for and retailers. Traditional time-series models like ARIMA improving demand forecasting. These techniques are and SARIMA have been widely used for demand capable of learning complex, non-linear relationships in forecasting, but their limitations in handling complex, large datasets, making them well-suited for supply chain non-linear relationships and incorporating external forecasting. Among these, deep learning models, factors such as promotions and weather events have led particularly Long Short-Term Memory (LSTM) networks, to the exploration of machine learning (ML) and deep have demonstrated significant promise in capturing learning (DL) techniques. This study evaluates and temporal dependencies and learning from sequential compares the performance of AI-driven demand data. The goal of this research is to evaluate the forecasting models, including ARIMA, SARIMA, Random effectiveness of AI-driven demand forecasting models, Forest (RF), Gradient Boosting Machines (GBM), and focusing on multi-echelon supply chains, and compare Long Short-Term Memory (LSTM) networks. The results their performance against traditional forecasting demonstrate that the LSTM model outperforms methods. traditional methods and other machine learning This paper presents a comprehensive methodology for algorithms in terms of accuracy, as measured by lower AI-driven demand forecasting, focusing on the MAE, RMSE, and MAPE values across all echelons of the integration of machine learning techniques, such as supply chain (retailer, wholesaler, and manufacturer). Random Forests, Gradient Boosting Machines, and The superior performance of LSTM highlights its ability LSTMs, for improving demand predictions at different to capture long-term dependencies and handle the supply chain echelons. The subsequent sections complexity of multi-echelon supply chains. This study describe the dataset collection, data preprocessing, provides valuable insights into the effectiveness of AI- model development, and evaluation processes, driven forecasting models for real-world supply chain followed by a comparative analysis of the results. applications, particularly in managing dynamic demand Literature Review patterns and optimizing operations. Demand forecasting has been a widely studied topic in Keywords: Demand forecasting, multi-echelon supply supply chain management (SCM). Traditional chain, Machine learning, Deep learning, Long Short- forecasting techniques, such as time-series analysis, Term Memory (LSTM), Random Forest, Gradient have long been used in practice. The AutoRegressive Boosting Machines, ARIMA, SARIMA, Supply chain Integrated Moving Average (ARIMA) model is one of the optimization, Forecasting accuracy. most well-known and widely applied methods for