An integrated decision analytic framework of machine learning with multi-criteria decision making for multi-attribute inventory classification

•A multi-criteria inventory classification method was developed.•Machine learning algorithms are integrated with multi-criteria decision making.•A case study at an automotive company validates the model with its high accuracy.•The proposed method yields significantly better results than others in literature.•It is flexibly applicable to other multi-criteria inventory classification cases.

The purpose of this study is to develop a hybrid methodology that integrates machine learning algorithms with multi-criteria decision making (MCDM) techniques to effectively conduct multi-attribute inventory analysis. In the proposed methodology, first, ABC analyses using three different MCDM methods (i.e. simple-additive weighting, analytical hierarchy process, and VIKOR) are employed to determine the appropriate class for each of the inventory items. Following this, naïve Bayes, Bayesian network, artificial neural network (ANN), and support vector machine (SVM) algorithms are implemented to predict classes of initially determined stock items. Finally, the detailed prediction performance metrics of algorithms for each method are determined. The comprehensive case study executed at a large-scale automotive company revealed that the best classification accuracy is achieved by SVMs. The results also revealed that Bayesian networks, SVMs and ANNs are all capable of successfully dealing with the unbalanced data problems associated with Pareto distribution, and each of these algorithms performed well against all examined measures, thus validating the fact that machine learning algorithms are highly applicable to inventory classification problems. Therefore, this study presents uniqueness in that it is the first and foremost of its kind to effectively combine MCDM methods with machine learning algorithms in multi-attribute inventory classification and is practically applicable in various inventory settings. Furthermore, this study also provides a comprehensive chronological overview of the existing literature of machine learning methods within inventory classification problems.