MRO Inventory Demand Forecast Using Support Vector Machine – A Case Study

Abstract

Today’s world is living in the age of digital transformation, the so-called Industry 4.0, in which technological advances have revolutionized the decision-making process in supply chain management. In this domain, inventory management can represent 50% of all organizational costs, and still a challenging task to keep the trade-off between maintaining inventory levels as low as possible, meeting clients’ demands, and maintaining satisfactory service levels. Forecasting the MRO inventory demand is even a more difficult task. To address this problem, machine learning (ML) applications, which deal well with nonlinear data, can predict irregular and intermittent demand with better accuracy than traditional approaches. This study employed the Support Vector Machine model to predict maintenance parts demand in a railroad logistic operator case study. This technique can deal with the nonlinear data encompassed by demand variations, avoid overfitting, and produce very accurate classifiers. Results indicated a considerable improvement in the demand forecast performance of the selected SKUs. This model can enhance the reliability of the purchasing and stock maintenance process and generate financial gains by reducing the need for large volumes of safety stock and greater assertiveness in meeting internal demands. It also contributes by showing a real case with an ML approach to predict inventory demands.