Machine Learning based Combinatorial Test Cases Ordering Approach

Combinatorial testing is an efficient test method, which can achieve high test coverage with as few test cases as possible. However, there are a large amount of test cases of combinatorial testing in industrial practice. If all the test cases are applied to executing, it takes a vast time and cost. How to select a subset of test cases which can guarantee failure detection rate is a common problem. In this paper, we introduce a novel technique for test case prioritization of combinatorial testing based on supervised machine learning. Our approach considers the test results of a small t-way covering array and the machine learning algorithm SVM is used to learn the test results first. Then SVM is used to predict a large t-way covering array. The test cases in the large t-way covering array are ordered according to the predicted results. The test cases which can lead failures in system under tests are ordered ahead. They own the priority. A subset of the ordered covering array which is selected from the start of the covering array can replace the whole covering array with time and cost saved reasonably. Our approach is evaluated by means of comparing the covering arrays which are ordered by SVM and the random ordering. Our results imply that our technique improves the failure detection rate significantly.