Improving Cost Visibility in Agile (Software) Projects Through a Fuzzy AHP–Q-Learning Prioritization Framework

Abstract

The evolution of project management methodologies has significantly impacted software development, particularly due to its adaptable and cyclical approach. Nonetheless, it introduces distinct difficulties in effectively recognizing and prioritizing cost overheads. Conventional cost estimation methods frequently prove inadequate in agile settings because of their rigid assumptions and failure to accommodate uncertainty and evolving conditions. This paper presents an enhanced classification framework aimed at prioritizing Agile Cost Overhead. It integrates the Fuzzy Analytic Hierarchy Process (Fuzzy AHP) with Q-Learning Optimization to effectively tackle existing limitations. The Fuzzy AHP method facilitates a systematic approach to prioritizing cost factors by leveraging expert insights and linguistic preferences, effectively addressing the inherent ambiguity present in agile processes. In addition to this, Q-Learning—an adaptive reinforcement learning method—enhances prioritization by learning from both historical and real-time data, consistently improving decisions through feedback driven by rewards. The hybrid framework is proposed by implementing a Voting Classifier that combines the predictions of Random Forest (RF) and Support Vector Machine (SVM), where the soft voting mechanism is applied for taking the final predictions. It categorizes overhead elements based on their impact (high, medium, and low) and adapts in real-time to project conditions, enhancing cost visibility and facilitating proactive decision-making. This model assists project managers in pinpointing essential cost factors and optimizing resource allocation. Assessment via simulated agile scenarios reveals improved prioritization precision and flexibility in contrast to traditional approaches. The proposed framework effectively connects expert intuition with intelligent learning, providing a scalable, data-driven solution for managing cost overhead in both software and non-software domains in an agile manner.