Context:

User interface testing validates the correctness of an application through visual cues and interactive events emitted in real-world usages. Performing user interface tests is a time-consuming process, and thus, many studies have focused on prioritizing test cases to help maintain the effectiveness of testing while reducing the need for full execution.

Objective:

This paper describes a novel test prioritization method called RLTCP whose goal is to maximize the number of test faults detected while reducing the amount of test.

Methods:

We define a weighted coverage graph to model the underlying association among test cases for the user interface testing. Our method combines Reinforcement Learning (RL) and the coverage graph to prioritize test cases. While RL is found to be suitable for rapidly changing projects with abundant historical data, the coverage graph considers in-depth the event-based aspects of user interface testing and provides a fine-grained level at which the RL system can gain more insights into individual test cases.

Results:

We experiment and assess the proposed method using nine data sets obtained from two mature web applications, finding that the method outperforms the six, including the state-of-the-art, methods.

Conclusions:

The use of both reinforcement learning and the underlying structure of user interface tests modeled via the coverage has the potential to improve the performance of test prioritization methods. Our study also shows the benefit of using the coverage graph to gain insights into test cases, their relationship and execution history.

中文翻译：

---

RLTCP：一种增强学习方法，用于对自动用户界面测试进行优先级排序

语境：

用户界面测试通过在实际使用中发出的视觉提示和交互式事件来验证应用程序的正确性。执行用户界面测试是一个耗时的过程，因此，许多研究都集中于对测试用例进行优先级排序，以帮助维持测试的有效性，同时减少对完整执行的需求。

客观的：

本文介绍了一种称为RLTCP的新型测试优先级排序方法，其目的是在减少测试数量的同时，最大化检测到的测试故障的数量。

方法：

我们定义了一个加权覆盖图，以对用于用户界面测试的测试用例之间的基础关联进行建模。我们的方法结合了强化学习（RL）和覆盖图来确定测试案例的优先级。尽管发现RL适用于具有丰富历史数据的快速变化的项目，但是覆盖图仔细考虑了用户界面测试中基于事件的方面，并提供了一个细粒度的层次，在该层次上RL系统可以获得对个人的更多见解测试用例。

结果：

我们使用从两个成熟的Web应用程序获得的九个数据集进行实验和评估，并发现该方法优于六个方法，包括最新方法。

结论：

通过覆盖范围使用强化学习和用户界面测试的基础结构，可以提高测试优先级排序方法的性能。我们的研究还显示了使用覆盖图来深入了解测试用例，它们之间的关系和执行历史的好处。