An Ensemble of Optimal Trees for Software Development Effort Estimation
| Autor: | Namir Abdelwahed, Zakrani abdelali, Moutachaouik Hicham |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Accurate estimation
Computer science business.industry Software development Software development effort estimation 020207 software engineering 02 engineering and technology Machine learning computer.software_genre Ensemble learning Regression Random forest Empirical research 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Artificial intelligence business computer |
| Zdroj: | Smart Data and Computational Intelligence ISBN: 9783030119133 |
| DOI: | 10.1007/978-3-030-11914-0_6 |
| Popis: | Accurate estimation of software development effort plays a pivotal role in managing and controlling the software development projects more efficiently and effectively. Several software development effort estimation (SDEE) models have been proposed in the literature including machine learning techniques. However, none of these models proved to be powerful in all situation and their performance varies from one dataset to another. To overcome the weaknesses of single estimation techniques, the ensemble methods have been recently employed and evaluated in SDEE. In this paper, we have developed an ensemble of optimal trees for software development effort estimation. We have conducted an empirical study to evaluate and compare the performance of this optimal trees ensemble using five popular datasets and the 30% hold-out validation method. The results show that the proposed ensemble outperforms regression trees and random forest models in terms of MMRE, MdMRE and Pred(0.25) in all datasets used in this paper. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|