An Ensemble Feature Optimization and Xtream Learning for High-Accuracy Software Reliability Modeling

Abstract

Software reliability is the domain that delivers accurate, reliable, and high-quality software by estimating faults early in development. Many existing reliability approaches have failed to estimate faults, errors, and expectations in software development. The main reasons are advanced programming languages, high-level features, and complex connections among several software metrics, which have a significant impact on software development performance. These issues are addressed by presenting the Ensemble Feature Optimization and Xtream Learning (EFXLM) framework for high-accuracy software reliability modeling. The proposed EFXLM is an integrated model that combines the Whale Optimization Algorithm (WOA) to estimate failure rate, using optimized tuning parameters to achieve accurate defect detection. To improve WOA performance, a Deep Extreme Learning Machine (DELM) is proposed to reduce failure rates by leveraging multiple hidden layers. The experiments are conducted on two benchmark datasets from the NASA and PROMISE repositories to evaluate the algorithms' performance. Performance can be measured as the failure prediction rate in binary classification.