Empirical investigation of hyperparameter optimization for software defect count prediction
作者：
Highlights：
• Examine effect of hyperparameters on regression techniques for software defect count prediction.
• Analysis the impact of hyperparameter tuning over 13 popularly-used regression techniques.
• Parameter optimization techniques perform better concerning the default parameter setting.
• Grid-search performs better than random-search for software defect count prediction.
• Parameter tuning changes the conclusion that which learners are better than others.
摘要
•Examine effect of hyperparameters on regression techniques for software defect count prediction.•Analysis the impact of hyperparameter tuning over 13 popularly-used regression techniques.•Parameter optimization techniques perform better concerning the default parameter setting.•Grid-search performs better than random-search for software defect count prediction.•Parameter tuning changes the conclusion that which learners are better than others.
论文关键词：Defect count prediction,Hyperparameter tuning,Machine learning
论文评审过程：Received 11 April 2020, Revised 28 September 2021, Accepted 8 November 2021, Available online 22 November 2021, Version of Record 7 December 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.116217
