Innovative applications of artificial intelligence in software testing

Monday, 12 July 2021

As artificial intelligence (AI) evolves, the scope of its applications continuously increases. Many state-of-the-art features like intelligent voice assistants, chatbots, and optical recognition are already part of our daily life. So much so, that many take these for granted. Similar developments are happening in the software development life cycle (SDLC). Innovation in this area is spreading fast and as new AI capabilities mature, these features will become normalized working practices moving forward. Breaking the SDLC down, the testing phase emerges as an immediate candidate where AI can have a significant impact.

Applying AI in software testing

Certain software testing automation tools can already process mundane and repetitive testing tasks in areas where human intervention adds no incremental value. However, the scope of these tools still has many limitations. AI can overcome these shortcomings to enable more automation, shorten testing cycles, accelerate decision-making, and deliver more ROI.

Concretely, there are five areas where AI is generating a step-change in the software testing life cycle.

1. Defect prediction and prevention

Deep learning-based analysis engines significantly improve the accuracy of defect prediction and prevention models. Historical statistics help identify high probability risk areas where defects can occur. Data on different types of common defect patterns that affect features, testing, or reporting, for example, are fed into analysis models to learn about their recurrence and accurately predict and prevent their appearance. Data on residual errors, like during User Acceptance Testing, fed into the models identify the gaps in testing and generate test cases automatically to cover those areas. In addition, customer testing patterns can be identified and test cases following those patterns generated. Once trained, the model can identify associated errors of a particular defect. Based on relativity data, a bug occurrence or fix can auto-generate test scenarios of impacted areas to enhance testing coverage.

2. Self-healing automation scripts

A common challenge software testers face in User Interface (UI) automation is the stability and maintenance of test scripts. AI technologies, particularly machine learning, can be used to extract details of UI elements like Document Object Models, attributes, and locators during each run, and map them with the actions performed during UI tests. When there is change in UI elements, AI can map these with the

respective elements in the model and generate locators automatically. This makes the locators smart and self-adjustable to changes. AI heuristics autonomously update testing models based on changes or modifications in the UI elements, rendering the tests fail-proof.

3. Auto-prioritization and selective test execution

Most automation tools either run all tests or a predetermined set since they do not know which tests to choose and execute dynamically. AI bots can analyze assigned test coverage and other metrics, like recent code changes or function criticality and stability among many others, to decide which specific tests to run. This reduces testing time by avoiding unwanted cases while ensuring testing coverage. Continuous integration and deployment (CI/CD) pipelines are major beneficiaries of this use case.

4. UI-driven API testing

Automated API tests are predominantly designed using set patterns that often do not cover the many variations that tend to arise during UI usage. Machine learning and AI can analyze these variations and detect patterns through API calls during the UI testing flows. This data is then applied to self-learning models which become increasingly comprehensive as they evolve over time, leading to stable, reusable, automated and scalable API tests.

5. Requirements-led Automatic Test Case Generation

A prerequisite for testing is establishing the parameters and criteria that need to be analyzed. Once the objectives are set, test cases or scenarios are developed. Currently. the quality of these cases highly depends on human intelligence. Yet, AI-driven natural language processing and machine learning can identify and execute these tests autonomously, directly from set requirements. AI can identify any changes in requirements or features that impact existing test cases and update them to cover the latest test scope. This frees testing engineers from spending time and resources on activities that can be automated, and instead focus on areas that demand higher levels of human intelligence and intervention.

High-quality software, faster

Applying AI techniques in various stages of the testing life cycle complements the tools that are currently available and enhances their value. Machine learning augments AI by continuously refining and enriching the information required for smart decision-making. Research in AI applications in the software testing life cycle is growing rapidly. Its goal – to enhance the efficiency and effectiveness of testing to create high-quality software, faster.

Find this content interesting? Subscribe to our blog for further updates!

Author Info

Rahul is a Senior Test Manager with over 17 years of experience in delivery, account management, and quality assurance. With an acute knowledge of quality process management, he focuses on bridging the gap between business performance and technology, tailoring test strategy based on architectural style and customer business processes. His expertise include developing continuous testing pipelines as part of automation and DevOPS initiatives, creating test solutions for web and mobile apps, digital products testing, data analytics testing, A/B testing, and PCI testing. As part of IBS Software's innovation and research in digital testing, he is working on ways to apply artificial intelligence techniques in testing life cycle. Rahul holds post-graduate degrees in electronic science and international business, and provides thought leadership to develop, implement, and evolve quality assurance processes, methodologies and best practices.

Tags:

Comment for this post has been locked by admin.

Comments

Innovation