Machine Learning System for Automated Testing

The evolution of the systems' complexity grew exponentially in the last years. The security and safety topics became more important than ever in the critical systems, and currently no end-user accepts any product without clear traceability for ensuring robustness to errors and external attacks. To be able to offer this kind of products, a high amount of effort must be invested in testing topics. Even that much part of the testing can be done automatically using automated test sequences, it is critical from the timing point of view to find as many errors as possible in the first hours/days of the testing time slot. The current paper presents a solution based on machine learning which decides the order of the tests, based on learned patterns: it analyses which functionalities are more prone to errors, and it generates the test sequence which needs to be executed at each step, in a recursive manner.