K -means clustering combined with principal component analysis for material profiling in automotive supply chains

At a time where available data is rapidly increasing in both volume and variety, descriptive data mining (DM) can be an important tool to support meaningful decision-making processes in dynamic supply chain (SC) contexts. Up until now, however, scarce attention has been given to the application of DM techniques in the field of inventory management. Here, we take advantage of descriptive DM to detect and grasp important patterns among several features that coexist in a real-world automotive SC. Principal component analysis (PCA) is employed to analyse and understand the interrelations between ten quantitative and dependent variables in a multi-item/multi-supplier environment. Afterwards, the principal component scores are characterised via a K-means clustering, allowing us to classify the samples into four clusters and to derive different profiles for the multiple inventory items. This work provides evidence that descriptive DM contributes to find interesting feature-patterns, resulting in the identification of important risk profiles that may effectively leverage inventory management for improved SC performance. [Received: 5 April 2019; Revised: 1 December 2019; Revised: 22 January 2020; Accepted: 21 April 2020]