Software Reliability Prediction with Weibull Failure Models

Abstract

Software reliability prediction helps testing teams estimate whether a system is becoming stable enough for release by analyzing observed failure-time behavior. This article presents a reliability prediction framework using Exponential and Weibull failure models applied to software testing intervals. The framework prepares confirmed failure records, calculates cumulative testing time and inter-failure time, estimates Exponential failure rate and Weibull shape and scale parameters, and compares reliability values across testing stages. The study shows that the Exponential model provides a simple constant-rate baseline, while the Weibull model gives deeper insight when failure intensity changes during debugging and retesting. The Weibull shape parameter is especially useful for identifying whether failure occurrence is decreasing, stable, or increasing over time. The proposed approach supports release-readiness assessment, testing continuation decisions, and reliability trend interpretation by converting raw failure logs into quantitative reliability evidence.