Coverage-based, prioritized testing using neural network clustering

Abstract

Graph-based algorithms are commonly used to automatically generate test cases for coverage-oriented testing of software systems. Because of time and cost constraints, the entire set of test cases generated by those algorithms cannot be run. It is then essential to prioritize the test cases in sense of a ranking, i.e., to order them according to their significance which usually is given by several attributes of relevant events entailed. This paper suggests unsupervised neural network clustering of test cases for forming preference groups, where adaptive competitive learning algorithm is applied for training the neural network used. A case study demonstrates and validates the approach.