Early Fault Detection and Quality Prediction in Software Development: A Fuzzy C-Means Clustering Approach with Attribute Selection and Evaluation

Abstract

Modern systems heavily rely on software-based systems, where software quality and reliability have emerged as significant concerns during the software development process. Developing software without any defects have proven challenging, as faults are inherent in software modules and can lead to failures in executable projects. Detecting fault-prone software components at an early stage becomes crucial for verification experts to focus their efforts effectively. This research proposes a novel approach using Fuzzy C-Means clustering for software fault and quality prediction. The approach employs attribute selection to identify crucial attributes and reduce their number, followed by attribute evaluation to assign weights. Through a ranking process, the most significant attributes are identified. Subsequently, Fuzzy C-Means clustering is utilized to group data points, where each cluster maintains a centroid value. A metric threshold is employed to determine cluster quality, with clusters having centroid values exceeding the threshold labeled as faulty, while those below the threshold are labeled as non-faulty. This methodology contributes to early fault detection and quality prediction in software development.