DESIGN HYBRID META-HEURISTIC APPROACHES FOR IMPROVED RELIABILITY OPTIMIZATION

Software Reliability Growth Models (SRGMs) are essential for assessing software dependability. The reliability evaluation process involves two key steps: model building and variable estimation, with this study focusing on the latter. Traditional methods like Least Squares Estimation (LSE) and Maximum Likelihood Estimation (MLE) were widely used for parameter estimation. However, these methods have limitations, increasing interest in metaheuristic optimization techniques. Metaheuristics overcome traditional drawbacks by employing strategies such as search field exploration and neighbourhood exclusion. This study evaluates four metaheuristic methods for SRGM variable estimation: Gravitational Search Algorithm (GSA), Sine-Cosine Algorithm (SCA), Grey-Wolf Optimizer (GWO), and Regenerative Genetic Algorithm (RGA). These methods were tested on three real loss datasets generated by four well-known SRGMs. The estimated variables using metaheuristic approaches closely align with those derived from LSE, demonstrating their accuracy. Results showed that RGA and GWO outperformed other techniques, offering superior parameter estimation capabilities. Additionally, RGA and GWO showed better integration and R2 dispersion values, making them more effective for practical failure data analysis. This research highlights the potential of RGA and GWO as reliable tools for SRGM parameter estimation, indicating their suitability for handling complex optimization challenges in software reliability studies.