Romina Shafieha - Department of Business Analytics & Operations, Surrey Business School, Guildford, UK
Mehdi Toloo - Department of Business Analytics & Operations, Surrey Business School, Guildford, UK

This article employs machine learning techniques with the aim of addressing a customer classification problem in the steel industry. The steel industry stands as a foundational pillar in any civilisation's development. The data used for classification is provided by an Iranian steel trading company which includes the purchase history of the customers of the company for the first three months of the year ۲۰۲۲. The company was previously using an expert-based grading system to segment its customers. We utilise the K-means model, a well-known unsupervised classification model, for classifying the customers who made at least one purchase in that period. We investigate both log modulus transformation and robust scaling methods to find the one that resulted in a more balanced impact of the variables, and subsequently better performance of the classification model. The number of clusters is determined using the elbow method, a popular technique for identifying the optimal number of clusters. The outcome obtained from our approach was compared with the expert-based grading system. The former and latter methods lead to three and eight classes of customers, respectively. This study reveals that our suggested data-based classification is more evenly, manageably, and reliably distributed than the traditional expert-based one.

Machine Learning, Classification, k-means, Steel industry