ENHANCED SOFTWARE RISK ANALYSIS IN REQUIREMENTS ENGINEERING USING FEPP AND MULTI-CLASS RULE LEARNING

Abstract

Software development projects often face challenges arising from incomplete, ambiguous, or evolving requirements, leading to increased risks during the early stages of development. Accurate prediction of potential risks in the requirements engineering (RE) phase is essential to ensure project stability, reduce rework, and improve overall software quality. This research introduces FEPP (Feature Extraction and Prediction Platform), an intelligent framework designed to enhance software risk analysis through innovative rule extraction and multi-class learning integration. The proposed FEPP model employs a hybrid approach that combines statistical analysis, machine learning, and explainable rule-based reasoning to identify and classify requirement-related risks into multiple categories, such as ambiguity, inconsistency, and incompleteness. The framework begins by performing comprehensive feature extraction from requirements documents, capturing both linguistic and semantic characteristics. These features are then processed through a multi-class classification pipeline using algorithms such as Random Forest, Gradient Boosting, and Support Vector Machines, integrated through an ensemble mechanism to improve predictive accuracy and model stability. A key component of FEPP is its rule extraction module, which enhances interpretability by generating human-understandable rules that explain how risks are identified and categorized. This not only aids in transparency but also allows software engineers and project managers to take proactive measures for risk mitigation. Experimental evaluations on real-world software requirement datasets demonstrate that the proposed framework significantly outperforms traditional binary classification and heuristic methods in accuracy, precision, and recall. By combining multi-class classification with explainable rule extraction, FEPP provides a robust, interpretable, and scalable solution for early-stage risk prediction in software development. The framework lays the foundation for integrating artificial intelligence into the requirements engineering process, ensuring better decision-making, improved software reliability, and reduced project failures.