A time-variant fault detection software reliability model

Abstract

AbstractIn this work, we propose a time-variant software reliability model (SRM)which considers the fault detection and the highest number of faults in software. The time-variant genetic algorithm process is implemented for the assessment of the SRM parameters. The proposed model works upon a non-homogeneous Poisson process (NHPP) and incorporates fault dependent detection and software failure intensity and the un-removed error in the software. We had considered programmers proficiency, software complexity, organization hierarchy, and perfect debugging as the determining factors for SRM. The dataset collected from 74 software projects was experimented with to establish and validate the proposed software reliability model's better fit. Data is collected over a period, which is initiated with the start of the project and is continuously monitored until its completion. Several parameters are analyzed, and a collection of 115 attributes are given with 11 different time frames in terms of product and process characteristics. A total of 383 persons were involved in software design, where the issue count total is 255. The proposed time-variant fault detection SRM is implemented in Jira and is also compared with the existing reliability model presented in the literature. It is observed that the proposed fault detection SRM works better in terms of different parameters like mean square error (MSE), root mean square error (RMSE), and r-squared (R2). The work is carried out, ensuring time-varying fault detection, which is measured by considering response count, coding and non-coding deliverables, and the number of bugs in the software. We considered the programmer's proficiency, software complexity, organization hierarchy, and perfect debugging as the determining factors for presenting the software reliability model. The proposed Software reliability model shows improvement over existing algorithms as the residual errors are reduced, and prediction accuracy is high in terms of cumulative fault detection.