A novel ML-MCDM-based decision support system for evaluating autonomous vehicle integration scenarios in Geneva’s public transportation

Abstract This paper proposes a novel decision-support system (DSS) to assist decision-makers in the ULTIMO project with integrating Autonomous Vehicles (AVs) in Geneva, Switzerland. Specifically, it aids in selecting the best scenario for incorporating AVs into Geneva’s public transportation system. The proposed DSS is architected on a combined integrated framework that includes a machine learning (ML) algorithm, random forest (RF) algorithm, and three novel multi-criteria decision-making (MCDM) algorithms: (1) Modified E-ARWEN (ME-ARWEN) for selecting the best scenario with high sensitivity; (2) Compromiser—Positive, Neutral, Negative (Compromiser-PNN) for extracting weights from stakeholders, considering their preferences and potential conflicts; and (3) Collective Weight Processor (CWP) for deriving weights from expert opinions. Besides the main objective, this article also aims to: (1) Address the gap in practical DSS software within AV-related studies by providing Python codes of the DSS; (2) Develop a highly sensitive and comprehensive MCDM framework to address the project’s needs; and (3) Employ Artificial Intelligence within the DSS to optimize outputs. By the application of the proposed DSS, four scenarios were evaluated: (1) Full integration of AVs; (2) Partial integration; (3) Pilot project in limited areas; and (4) Delayed integration. The analysis identified partial integration as the best scenario for integrating AVs. Furthermore, comprehensive analyses conducted to validate the DSS outputs demonstrated the reliability of the results.