AN ADAPTIVE GA-OPTIMIZED AUTOMATED X-LEARNING MODEL FOR INTEGRATED SOFTWARE RELIABILITY PREDICTION

Abstract

Software reliability is a significant challenge for software companies today, enabling the development of high-quality, stable, and sustainable software systems. Many existing approaches lack accuracy, are poor in prediction, and fail to handle the highly complex software bugs. To overcome these challenges, we have presented an Adaptive Genetic Algorithm (GA)-Optimized Automated X-Learning Model (OA-XLM) for integrated software reliability prediction. The main aim of this model is to predict and detect the accuracy bugs (software threats) from the latest software datasets. The proposed approach combines an adaptive genetic algorithm (AGA) to find an accurate Learning rate, an in-depth selection of intelligent features, and hyper parameters optimization. Here, the automated X-learning model (XLM), also known as Extreme-Learning, performs bug prediction and failure rate estimation. The XLM integrates multiple learning layers to capture both linear and nonlinear relationships in software data. Finally, the XLM also classifies the software bugs with binary classification. This model was primarily developed to adapt dynamically across complex software projects. In this work, two benchmark datasets are used, namely the NASA and PROMISE repositories. The results show that the proposed approach achieved high prediction and classification rates.