Scholarlink

<p>Accurate estimation of software testing efforts is challenging, as changes occur frequently throughout the project life cycle. Thus, employing machine learning (ML) techniques helps make good estimates, but it cannot ensure that these estimates are based on correct and unbiased data, especially when using real-world data. Modern research using ML techniques has to focus on accuracy and achieving a certain quality level of the input data (<i>i.e.,</i> interpretability) to be trained and tested. In this study, we propose an enhancement test effort estimation (TEE) model that integrates ML with ontology-based data interpretability improvements. Three main steps make up the methodology: (i) gathering real-world data including change request (CR) for testing; (ii) improving it using an ontological model and linking each test case (TC) to its functional size (FS) using the COSMIC Functional Size Measurement (FSM) method to improve interpretability; and (iii) developing an ML-based TEE model that incorporates two algorithms (Support Vector Regression (SVR) and M5P). Experimental results demonstrate that improving data interpretability significantly improves the accuracy of effort estimation. Thus, after applying our methodology, the SVR-enhancement TEE model achieves a 62.5% reduction in MAE and a 54.5% reduction in RMSE, while the M5P-enhancement TEE model shows a 58.3% reduction in MAE and a 50% reduction in RMSE. The findings show that high-quality data plays a key role in improving enhancement TEE-based ML models.</p>

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

Ontology inference for data quality improvement: a support vector regression-based test effort estimation model

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Zaineb Sakhrawi,
  • Taher Labidi,
  • Asma Sellami,
  • Nadia Bouassida

摘要

Accurate estimation of software testing efforts is challenging, as changes occur frequently throughout the project life cycle. Thus, employing machine learning (ML) techniques helps make good estimates, but it cannot ensure that these estimates are based on correct and unbiased data, especially when using real-world data. Modern research using ML techniques has to focus on accuracy and achieving a certain quality level of the input data (i.e., interpretability) to be trained and tested. In this study, we propose an enhancement test effort estimation (TEE) model that integrates ML with ontology-based data interpretability improvements. Three main steps make up the methodology: (i) gathering real-world data including change request (CR) for testing; (ii) improving it using an ontological model and linking each test case (TC) to its functional size (FS) using the COSMIC Functional Size Measurement (FSM) method to improve interpretability; and (iii) developing an ML-based TEE model that incorporates two algorithms (Support Vector Regression (SVR) and M5P). Experimental results demonstrate that improving data interpretability significantly improves the accuracy of effort estimation. Thus, after applying our methodology, the SVR-enhancement TEE model achieves a 62.5% reduction in MAE and a 54.5% reduction in RMSE, while the M5P-enhancement TEE model shows a 58.3% reduction in MAE and a 50% reduction in RMSE. The findings show that high-quality data plays a key role in improving enhancement TEE-based ML models.