| 摘 要: 在持续集成环境下,针对大多数测试用例的优先排序方法难以有效选择和利用历史信息的问题,提出了一种基于历史信息的自适应动态划分机制(ADPM)。首先,该机制为测试用例分配若干动态信息块,并根据测试用例潜在失效影响动态划分历史信息,实现测试用例潜在失效能力的度量,并以此生成历史信息动态特征;其次,结合轻量级深度学习框架预测测试用例优先级,调整测试用例执行顺序。为验证ADPM方法的有效性,在3个工业数据集上进行了实验,采用归一化平均故障检测百分比(NAPFD)作为主要评价指标。实验结果表明,使用ADPM后,测试用例序列的 NAPFD较使用固定长度历史信息平均增长了10个百分点。 |
| 关键词: 持续集成 回归测试 测试用例优先排序 历史信息选择 深度学习 |
|
中图分类号:
文献标识码: A
|
| 基金项目: 国家自然科学基金资助项目(61663005) |
|
| Research on Adaptive Dynamic Partitioning Mechanismin Continuous Integration Test Optimization |
|
WANG Yuanhuan, YANG Yang, GUO Liang
|
(School of Information Science and Engineering, Zhejiang Sci-Tech University, Hangzhou 310018, China)
202230705158@mails.zstu.edu.cn; yangyang0070@zstu.edu.cn; lguo@zstu.edu.cn
|
| Abstract: In the context of continuous integration, addressing the challenge that most test case prioritization methods struggle to effectively select and utilize historical information, an Adaptive Dynamic Partitioning Mechanism(ADPM) based on historical information is proposed. Firstly, this mechanism allocates several dynamic information blocks to test cases and dynamically partitions historical information based on the potential failure impact of test cases.This enables the measurement of the potential failure capability of test cases, thereby generating dynamic features of historical information. Secondly, the priority of test cases is predicted by integrating a lightweight deep learning framework, and the execution order of test cases is adjusted accordingly. To validate the effectiveness of the ADPM method, experiments were conducted on three industrial datasets, using the Normalized Average Percentage of Faults Detected (NAPFD) as the primary evaluation metric. The experimental results show that after applying ADPM, the NAPFD of the test case sequence increases by an average of 10 percentage points compared to using fixed-length historical information. |
| Keywords: continuous integration regression testing test case prioritization historical information selection deep learning |
