Moustafa F. Ali, Angie M. Ceniza-Canillo
Software development methodology selection remains a critical decision that significantly impacts project success rates, development efficiency, and team productivity, with traditional approaches relying heavily on expert judgment and static decision matrices that often fail to adapt to evolving project requirements and emerging methodologies. This study presents a novel knowledge-based expert system that integrates Multi-Criteria Decision Making (MCDM) techniques with Flow Entropy Sort algorithms, enhanced by Large Language Models (LLMs) for intelligent software methodology selection. We developed a hybrid system combining rule-based expert knowledge with LLM-powered natural language processing capabilities, employing Analytic Hierarchy Process (AHP) for criteria weighting, TOPSIS for alternative ranking, and Flow Entropy Sort for dynamic preference ordering, with a GPT-4 based module that processes project descriptions, extracts relevant features, and provides contextual recommendations. The system was evaluated using 150 real-world software projects across various domains, achieving 87.3% accuracy in methodology selection compared to expert recommendations, representing a 23% improvement over traditional rule-based systems, with response times averaging 2.4 seconds per query and user satisfaction scores reaching 4.2/5.0, while the LLM component successfully identified implicit project requirements in 78% of cases, leading to more informed decision-making. The integration of LLMs with traditional MCDM approaches demonstrates significant potential for automating complex software engineering decisions, providing explainable recommendations while maintaining high accuracy and user acceptance rates. © 2025 Copyright held by the owner/author(s).
Department of Computer, Information Sciences and Mathematics, University of San Carlos, Cebu, Philippines
