An Order Classification Model Using Machine Learning in Hybrid MTS/MTO Production Systems

Companies frequently utilized multiple sales channels to ensure sustainability. However, attempts to fulfill all incoming orders without sufficient consideration of operational constraints resulted in significant challenges in planning and production processes. Therefore, effective order management was considered essential for optimizing overall business performance. The strategic evaluation of sales orders reduced post-acceptance difficulties, thereby improving both customer satisfaction and operational efficiency. In this context, a machine-learning-based model for sales order classification was proposed for a hybrid production environment accommodating both Make-to-Stock (MTS) and Make-to-Order (MTO) strategies. Initially, the attributes employed in sales order evaluation were identified through an extensive literature review and subsequently refined using a heuristic approach to determine the most relevant classification features. Based on these attributes, sales orders were first clustered into three groups using the k-means algorithm to generate meaningful class labels. The labeled datasets were then utilized to train three supervised machine learning algorithms: Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN). The performance of these models was evaluated and compared, resulting in accuracy rates of 99.67%, 99.55%, and 99.49%, respectively. The findings demonstrated that the Random Forest algorithm achieved the highest classification performance.