Scholarlink

For a considerable time, the software development life cycle has been obstructed by the volatile factors associated with software cost estimation. The stakeholders involved have been severely impacted by the inaccuracies they cause during the estimating process. This can be alleviated by estimating the cost using machine learning algorithms, which significantly reduce process volatility and produce more accurate outcomes. Thus, implementing analogy-based estimation along with stacking using SVR, Ridge Regressor, K-nearest neighbours and Linear Regression in level-0 and SGD Regressor in level-1, on various datasets has given highly accurate results. The Metaheuristic and Analogy-based approach, for software cost estimation using ensemble learning, built with metaheuristic algorithm-based hyperparameter tuning that we propose is substantiated on 4 datasets Nasa93, China, Maxwell and ISBSG and evaluated using MAE, RMSE, R2, PRED, MMRE evaluation metrics. We find competent performance being displayed by our proposed model with ISBSG and China datasets displaying the most promising results.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybridizing Metaheuristics and Analogy-Based Methods with Ensemble Learning for Improved Software Cost Estimation

  • 章节

  • 发表日期：

    2025年

查看全文
  • Anupama Kaushik,
  • Kalpana Yadav,
  • Prabhjot Kaur,
  • Kavita Sheoran,
  • Nikhil Bhutani,
  • Ritvik Kapur,
  • Bhavesh Singh

摘要

For a considerable time, the software development life cycle has been obstructed by the volatile factors associated with software cost estimation. The stakeholders involved have been severely impacted by the inaccuracies they cause during the estimating process. This can be alleviated by estimating the cost using machine learning algorithms, which significantly reduce process volatility and produce more accurate outcomes. Thus, implementing analogy-based estimation along with stacking using SVR, Ridge Regressor, K-nearest neighbours and Linear Regression in level-0 and SGD Regressor in level-1, on various datasets has given highly accurate results. The Metaheuristic and Analogy-based approach, for software cost estimation using ensemble learning, built with metaheuristic algorithm-based hyperparameter tuning that we propose is substantiated on 4 datasets Nasa93, China, Maxwell and ISBSG and evaluated using MAE, RMSE, R2, PRED, MMRE evaluation metrics. We find competent performance being displayed by our proposed model with ISBSG and China datasets displaying the most promising results.