A multi-criteria decision-making and machine learning-basedapproach for enterprise resource planning (ERP) system selection inthe electric power distribution sector

This thesis aims to analyze which Enterprise Resource Planning (ERP) software alternative should be prioritized under limited resources within the scope of information technology (IT) investments of a large-scale organization operating in the electricity distribution sector, through an objective and multicriteria decision-making (MCDM) framework. ERP systems are strategic information systems that integrate core business functions such as production, accounting, human resources, and supply chain management. In this context, the ERP selection process evolves into a complex decision problem characterized by multiple criteria and alternatives. In this study, an integrated application of three distinct MCDM methods—Analytic Hierarchy Process (AHP), TOPSIS, and VIKOR—was employed. Based on literature and expert opinions, four main criteria (cost, technical capability, adaptability, and vendor factors) and a total of 17 sub-criteria were defined. Subsequently, four ERP alternatives (SAP, Oracle, MBS [Customer Information Systems], and a national open-source software, THOR) were evaluated within the context of Dicle Electricity Distribution Inc. According to the empirical findings, the criteria weights obtained through AHP revealed that adaptability was the most critical dimension in ERP evaluation, with a relative weight of 34.65%. Moreover, these criterion weights were integrated into supervised learning models such as the XGBoost algorithm, and a strong consistency was observed across methods. This approach introduced an innovative contribution by supporting traditional MCDM techniques with data-driven models. In terms of ERP alternative ranking, the analyses based on AHP, TOPSIS, and VIKOR all pointed to THOR as the most ideal solution. The convergence in rankings among different methods demonstrated the multidimensional nature of the decision problem and the methodological robustness of the proposed model. The decision support framework developed in this research not only offers a comparative analysis of MCDM methods at the academic level but also provides a systematic, applicable, and transparent roadmap for decision-makers in ERP investments. The adaptability of the model to sector-specific requirements further enhances decision quality, particularly in technically intensive sectors such as electricity distribution. The findings of this study contribute to improved decision quality and risk mitigation in high-cost IT investments such as ERP system selection.