An Explainable Deep Learning Model in Improving Test Case Prioritization for Continuous Integration Testing

Shankar Ramakrishnan, E. K Girisan

Abstract: Continuous Integration (CI) testing is a critical phase in current software industry. Test Case Prioritization (TCP) methods are introduced to enhance Regression Testing (RT) by ranking test cases for early developer feedback. Various Deep Learning (DL) models have been developed to improve TCP in CI environments. But, many struggle to simultaneously capture the structural relationships among test case features and the temporal dependencies across multiple testing cycles. Furthermore, these models often rely on large amounts of historical execution data, limiting their effectiveness in fast-paced and diverse CI scenarios. To address this issue, XCG-TCP, an eXplainable Convolutional Neural Network (CNN) - Gated Rectified Unit (GRU) is proposed for accurate TCP in CI. Initially, the collected test case data will be pre-processed using data cleaning, categorical encoding, and feature scaling to ensure balanced and consistent inputs. A Deep CNN (DCNN) extracts spatial and structural features from various test case attributes such as execution duration, previous results, status changes, execution flags, priority and code modification distance. These features are then fed into a Gated Recurrent Unit (GRU) to model the temporal dependencies and sequence patterns across regression cycles enabling effective identification of failing test cases. The integration of SHapley Additive exPlanations (SHAP) as an Explainable Artificial Intelligence (XAI) model enhances the CNN-GRU model by quantifying the influence of each input feature on the final TCP decision. This combination of these models enhances early fault detection, accelerates testing cycles and improves adaptability across varying CI environments. Experimental results demonstrate that XCG-TCP outperforms standard algorithms on industrial datasets.

Keywords: Continuous Integration, Test Case Prioritization, Convolutional Neural Network, Gated Rectified Unit, SHapley Additive exPlanations

How to Cite?: Shankar Ramakrishnan, E. K Girisan, "An Explainable Deep Learning Model in Improving Test Case Prioritization for Continuous Integration Testing", Volume 14 Issue 8, August 2025, International Journal of Science and Research (IJSR), Pages: 703-710, https://www.ijsr.net/getabstract.php?paperid=SR25813194634, DOI: https://dx.doi.org/10.21275/SR25813194634