Multi-Criteria Decision-Making-Based Model Selection Proposal in Artificial Learning Process

Abstract

In this study, selection of the best classification model in the artificial learning process is considered as a Multi-Criteria Decision-Making problem. In this direction, machine learning- based 10 classification models have been obtained and seven of them have been eliminated according to the different parameters. The classification model with the best performance among three remaining alternatives has been determined by ELECTRE I. According to the model performances obtained within the scope of the study, the best model among the three alternatives would be determined depending on the initiative of the researcher. However, with the proposed model, this process has been moved to a scientific basis and the best of the three models based on Extreme Learning Machine (ELM), Naïve Bayes, and Support Vector Machine has been clearly determined as ELM. The proposed model, unlike its counterparts in the literature, is far from a complex structure, is understandable and can support users of all levels.