Behaviour pattern-based model generation for model-based testing

This paper presents the concept of using behavioral pattern mining to generate models for model-based testing. These patterns are mined from observations captured from execution scenarios of the system under test, and the different patterns are combined in order to provide a suitable higher level model for model-based testing. The concept is first discussed on a general level, providing a basis for implementation of semi-automated model generation algorithms based on combinations of different behavioral patterns. This concept is illustrated by showing how to generate extended finite state machine models in a format suitable for model-based testing. The generated model is further validated by applying it for model-based testing of a real software component, where it reveals actual faults in the system under test. In addition to the benefits, the discovered limitations of the approach are discussed. Future work is discussed as potential means to address these limitations. (19 refs.)