Authors:
Suyash Shukla
and
Sandeep Kumar
Affiliation:
Department of Computer Science and Engineering, Indian Institute of Technology Roorkee, Roorkee, India
Keyword(s):
Software Effort Estimation, Machine Learning, Ensemble Models, ISBSG Dataset.
Abstract:
Software Effort Estimation (SEE) is the undertaking of precisely assessing the measure of effort needed to create software. A lot of exploration has already done in the field of SEE using Machine Learning (ML) strategies to deal with the deficiencies of traditional and parametric estimation methodologies and line up with present-day advancement. Nonetheless, generally due to questionable results and uncertain model development strategies, just a few or none of the methodologies can be utilized for deployment. This paper intends to enhance the procedure of SEE with the assistance of an ensemble based ML approach. So, in this study, a stacking ensemble-based approach has been proposed for SEE to deal with the previously mentioned issues. To accomplish this task an International Software Benchmarking Standards Group (ISBSG) dataset has been utilized along with some data preparation and cross-validation technique. The outcomes of the proposed approach are compared with Multi-Layer Percep
tron (MLP), Support Vector Machine (SVM), and Generalized Linear Model (GLM) to obtain the best performing model. From the results, it can be concluded that the ensemble model has produced fewer error estimates contrasted than other models. Lastly, we utilize the existing approaches as a benchmark and compared their results with the models utilized in this study.
(More)