Masterminds

Demand forecasting is known to be challenging due to phenomena such as the Bullwhip Effect (Lee et al., 1997a). Having worked in supply chain management for many years, I observed that the Bullwhip Effect is a universal issue across all industries. Companies, regardless of size, struggle with the amplification and distortion of demand information. My master’s thesis focused on leveraging AI, specifically machine learning, to enhance demand forecasting and optimize supply chains for the case company, a producer of durable IT devices. This thesis aimed to bridge two hot topics in the digital era — machine learning and demand forecasting — by providing a practical solution within a real business context. Effective demand forecasting and supply planning are crucial components of supply chain management. Inaccurate demand information often leads to suboptimal decisions, causing inventory imbalances and customer dissatisfaction. Many organizations struggle with business challenges due to inaccurate demand forecasts, resulting in inefficiencies, financial losses, and unhappy customers. Recent advancements in machine learning (ML) algorithms offer new tools to improve forecasting accuracy and maintain excellent performance for industrial demand. Since 2018, Machine Learning algorithms have consistently won competitions focused on retail demand forecasting. In my study, I employed an applied action research approach to diagnose the case company's challenges and offer viable solutions. Data collection primarily involved qualitative methods such as interviews, meetings, and internal document analysis, supplemented by quantitative data for model development. Four algorithms were used to build Machine Learning models using data extracted from the company's weekly demand reports: Linear Regression, Decision Tree, Recurrent Neural Network, and Support Vector Machine (as recommended by: Vandeput, 2023). After processing the data, performing feature engineering, and conducting training, testing, and validation, Linear Regression emerged as the most suitable algorithm based on both Machine Learning metrics and internal evaluations. Based on the thesis results, this approach was recommended for integration into the case company's existing demand forecasting and supply planning processes to support better decision-making. The steps in the approach are shown in Figure 1 below.   Figure 1. Demand Forecasting Process by Utilizing Machine Learning Model (Zhang 2024).   By leveraging AI technology, particularly machine learning, in demand forecasting and supply planning, organizations can vastly improve supply chain management. Enhanced forecast accuracy and optimized inventory levels lead to reduced inefficiencies, minimized financial losses, and heightened customer satisfaction. My academic journey in Metropolia University of Applied Sciences has equipped me with invaluable expertise in digitalization and data analytics. Notably, Power BI and Machine Learning provided direct insights for visualizing data, aiding in the selection of the most suitable machine learning model in my thesis.   About the author Yi Zhang is a graduate of Master´s degree program in Business Informatics. Yi is keen on new technologies and has set her personal goal to master Machine Learning for advancing in her professional area.   References Lee, H.L., Padmanabhan, V. & Whang, S. (1997 a). The Bullwhip Effect in Supply Chains. Sloan Management Review/Spring. Vandeput, N. (2023). Demand forecasting best practices. Manning Publications. Available from: https://learning.oreilly.com/library/view/demand-forecasting-best/9781633438095/?sso_link=yes&sso_link_from=metropolia-university Zhang, Y. (2024). Machine Learning Applied in Demand Forecasting and Supply Planning. Metropolia UAS, Master´s thesis, 99 pages. Available from: https://www.theseus.fi/handle/10024/857760