Sales forecasting in fashion retail chain: A case study in Vietnam

Abstract

Sales forecasting is the key to the success of a supply chain, especially in the fashion industry. Vietnam is a major apparel supplier for international brands and has a dynamic, fast-growing market. The lack of complete forecasting systems in such a market motivates our development of sales forecast procedures featuring quantifiable results and practical implementation. Both statistical and machine learning methods were tested using actual data. We found that Random Forest consistently yields the lowest error metrics, followed closely by XGBoost. Meanwhile, clustering did not provide conclusive evidence of improved accuracy. As the study’s data sources come entirely from one company, these procedures are applicable for other firms to deploy their data without external mining sources.