Prioritization of information technology investment projects inthe electricity distribution sector through the integration ofmulti-criteria decision making and machine learning

The primary objective of this thesis is to analyze, within a multi-criteria decision-making framework, which projects should be prioritized under limited resources in the scope of information technology (IT) investments of a large-scale institution operating in the electricity distribution sector in Turkey. Seven project alternatives proposed by the institution's technology department—Receivables Management System, Investment Management System, Fleet Management System, Subscriber Management System, Outage Management System, Demand Tracking System, and Inventory Management System—were evaluated according to the institution's strategic goals and operational needs through a multi-criteria assessment process. In the analysis, the Analytic Hierarchy Process (AHP) was employed to weight the decision criteria, while the AHP, PROMETHEE and VIKOR methods were applied to rank the project alternatives. The criteria were grouped under seven categories: project timeline, government support, investment budget, customer satisfaction, billing, collection, and loss-theft rate. The AHP results revealed that the highest weights were assigned to collection (0.3488), loss-theft (0.2505), and billing (0.2127), indicating that the institution prioritizes financial efficiency and operational control factors in its investment decisions. Although some ranking differences appeared between the AHP, PROMETHEE and VIKOR analyses, general tendencies were consistent across methods. Project 1 (Receivables Management System) was identified as the highestpriority project in all methods, notably due to its potential to enhance collection efficiency and improve financial processes. Project 5 (Outage Management System) and Project 4 (Subscriber Management System) also ranked among the top projects suitable for investment. Additionally, the study incorporates the modeling of evaluation results obtained from multi-criteria decision-making methods using the XGBoost algorithm, enabling a detailed analysis of the determinant influence of various criteria on project prioritization. This approach transcends the fixed weighting structures of traditional methods by facilitating the identification of dynamic criterion importances grounded in machine learning techniques.The findings demonstrate that multi-criteria decision-making methods can serve as effective tools in corporate project management and investment planning. Moreover, the integrated application of multiple methods increases the reliability of ranking results and provides decision-makers with a more balanced perspective. In this context, addressing multi-dimensional criteria—including not only technical and financial indicators but also strategic alignment and operational contribution—in IT investments is expected to yield positive long-term impacts on institutional performance.