Year: 2022
Type: article
Abstract: Growing reliance on software products in the world has put high demand on quality software.Software Bug Prediction is an exercise in enhancing the quality of software by identifying potential bugs or faults in software constructs during the pre-deployment testing phase.Various Machine Learning models have been built to predict faults based on metrics derived from the software.Feature selection and feature reduction (extraction) are two strategies to weed-out redundant and non-useful features thereby reducing the dimensionality of features.The present research study aims to devise an ML model to effectively reduce the dimensionality of the feature-set of data by using a combination of Genetic Algorithm (feature selection) and Feature Agglomeration (feature reduction) without significantly affecting model performance.Datasets containing 62 features are employed to test the model using four classifiers and three cross-validated metrics.Results show that the proposed model reduces the feature-set by more than 80 %.The application of the Kruskal Wallis H test shows the statistical similarity between the results of the proposed model and results without a proposed model.In a majority of test cases, the proposed model has improved the performance of base classifiers by extracting the most informative features.
Authors: Tamanna -, Om Prakash Sangwan
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4283217218
