Enhancing software reliability prediction using fuzzy AHP-based mathematical model and ANN integration

In recent years, intensive research in software engineering has been dedicated to predict the reliability of software systems. Multiple methodologies have been explored to evaluate and estimate the reliability of software systems. Software developers can streamline the process of creating new software by incorporating crucial elements such as reusability, component interaction, component dependency, component complexity, and failure. This study introduces two innovative models for software reliability assessment. The first model utilizes a mathematical framework, considering five key factors—reusability, component interaction, component dependency, component complexity, and failure—to construct a comprehensive mathematical model for software reliability assessment. The incorporation of Fuzzy Analytical Hierarchy Process is employed to determine the pertinent weights of these factors, thus contributing to a nuanced evaluation of software reliability. The second model leverages the power of Artificial Neural Network for software reliability assessment. Both proposed models exhibit superior reliability values when compared to various existing models. Notably, the average reliability scores computed across 100 programs for the proposed AHP-based model and ANN-based model are 0.438269 and 0.416136, respectively.