Decision Trees for Supervised Multi-criteria Inventory Classification

Abstract

A multi-criteria inventory classification (MCIC) approach based on supervised classifiers (i.e. decision trees and random forests) is proposed, whose training is performed on a sample of items that has been previously classified by exhaustively simulating a predefined inventory control system. The goal is to classify automatically the whole set of items, in line with the fourth industrial revolution challenges of increased integration of ICT into production management. A case study referring to intermittent demand patterns has been used for validating our proposal, and a comparison with a recent unsupervised MCIC approach has shown promising results.