XGBoost-Based Demand Forecasting in Supply Chain Management Using Machine Learning Algorithm

Abstract

Effective supply chain management (SCM) plays a significant role in enterprises seeking to mitigate risks, optimize overall productivity, and decrease costs while maintaining product quality and customer satisfaction. Addressing quality defects within the supply chain proactively assists in minimalizing returns, recalls, and rework, which results in increased profitability and considerable financial benefit. For the accurate delivery of products and collecting feedback for analysis, enterprises rely on SCM to satisfy customer needs. Machine learning (ML) approaches have witnessed a revolutionary milestone in SCM, facilitating more effective management of the supply chain. However, operational costs, model interpretability, and data quality remain to be a major challenges. Therefore, this paper presents XGBoost-based demand forecasting in SCM using an ML algorithm. At the initial stage, Z-score normalization servers as a data preprocessing to normalize the input data features into uniform scales. For the feature selection process, the genetic algorithm (GA) is exploited to identify the most descriptive variables. Then, the XGBoost model is employed for the demand forecasting process to accurately forecast future market trends. Finally, particle swarm optimization (PSO) involves hyperparameter tuning which allows the model to achieve optimal forecasting results. Experimental outcomes highlight that the proposed technique accomplishes superior forecasting performance compared to other existing approaches.