Using artificial intelligence to automate testing

Abstract

Automation of software testing is a critically important task in modern development aimed at improving the quality and reliability of products. Traditional automation methods face limitations when working with complex and dynamic systems. This study is aimed at evaluating the effectiveness of the use of artificial intelligence (AI) methods to overcome these limitations and improve the quality of automated software testing. An integrated approach was used in the work, including machine learning, neural networks and genetic algorithms. Based on an extensive set of test case data, models have been developed and trained to predict test results, generate optimal test data sets, and identify anomalies. The effectiveness of the developed methods was evaluated by comparing with traditional approaches on a number of metrics, including accuracy, completeness and F-measure. The use of AI methods has improved the accuracy of defect detection by 18% compared to the basic methods. The time spent on developing and executing tests has been reduced by 25%. Test coverage increased by 15% when using genetic algorithms to optimize test suites. The results demonstrate the significant potential of AI methods in software testing automation. Key advantages have been identified, including improved accuracy, reduced time costs and expanded test coverage. Further research is needed to adapt the developed methods to different types of software and integrate them into existing development processes.