Intelligent Test Execution: Risk‑Based Manual Case Recommendation, Parallel‑Coverage Traffic Selection, Smart Build, Priority‑Based Task Scheduling, and UI Automation Self‑Healing

In the previous chapter we introduced intelligent test input generation such as abnormal unit test creation, interface case generation, and action set generation. This chapter focuses on the intelligent practices for the test execution phase, which involves running generated test case and data sets manually or automatically.

01 Risk‑Based Manual Case Recommendation

Because code changes and environment factors make it unnecessary to run all test cases, selecting the most fault‑revealing cases is a key research area. Manual test case recommendation aims to select high‑coverage cases based on code changes, avoiding redundant recommendations caused by simple coverage‑based methods. The approach abstracts code into syntax trees, extracts 21 complexity metrics, and uses models such as Bayesian classifiers, SVM, KNN, logistic regression, or deep learning (LSTM, DNN) to predict defect‑prone code and recommend associated test cases. After deployment, case recommendation reduced the recommendation ratio from 50% to 20% and shortened regression cycles from three days to one day while increasing the proportion of bug‑finding cases.

02 Parallel‑Coverage Traffic Selection

During testing, large volumes of online traffic are used for diff, performance, and stress testing. The parallel‑coverage traffic selection method aims to find a minimal traffic set that covers the most test scenarios, improving coverage while reducing traffic volume. It consists of two steps: (1) initial log‑based filtering by device, region, user attributes, etc., and (2) greedy algorithm‑driven selection based on coverage analysis to achieve maximal scenario coverage with minimal traffic. This approach has been applied across multiple product lines, halving traffic volume while maintaining or improving coverage by up to 60%.

03 Smart Build

Smart build dynamically adjusts CI tasks based on change characteristics, enabling task pruning, skipping, cancellation, result reuse, self‑healing, and auto‑annotation. Typical scenarios include minor log or format changes, repeated execution of the same task, and redundant runs on branches versus trunk. Traditional full‑build approaches waste resources; smart build analyzes change features (e.g., changed lines, functions, call chains) and decides whether to execute, wait, restart, or cancel tasks, often using whitelist rules. Baidu’s smart‑build system, comprising strategy developers, plugins, and business users, now serves over 3,000 modules.

04 Priority‑Based Task Scheduling

This research addresses how to schedule test tasks under limited resources while preserving stability and fault‑detection capability. By constructing a priority queue based on task importance, waiting time, and resource demand, the system reduces average waiting time for critical tasks. Offline analysis of historical task durations and coverage informs optimal stop‑time prediction, while a real‑time stop‑decision model monitors execution metrics (duration, screenshot count, UI coverage change) to terminate low‑value tasks early. The result is a 10% reduction in execution time without degrading coverage, and a 12% time saving with a 10% coverage increase.

05 UI Automation Self‑Healing

Automated app test cases often encounter unexpected situations such as upgrade pop‑ups, slow page loads, or changed XPaths, causing failures and high maintenance costs. Three self‑healing techniques are employed: (1) abnormal pop‑up handling using object detection and text‑based classification to identify and dismiss pop‑ups; (2) atomic wait technology that leverages visual UI understanding and video frame analysis to intelligently adjust wait times; (3) generic case self‑healing that records successful locators (XPath, icons) and retries with historical elements upon failure. These methods achieved a 51% case self‑healing rate, significantly improving automation stability.

Recommended reading links are provided for deeper insights into Baidu’s intelligent testing initiatives.