SML4C: Fully Automatic Classification of State Machine Models for Model Inspection in Education

In the task-based learning of state machine modeling, understanding a large number of learner-created models is a very time-consuming task for instructors. Given that learner-created models with different descriptions often represent the same behavior, if such models can be classified correctly by behavioral similarity, then the instructors may efficiently create feedback with no oversights by inspecting one model per class. Such approach, in essence, allows learners to receive an early and beneficial feedback. However, the problem is that there is no existing method that can easily and automatically classify these learner-created state machine models in terms of behavioral similarity, which is our firsthand attempt in this study. We referred to our proposed tool as SMart-Learning for Classification (SML4C). It specifically measures the behavioral similarity of the state machine models based on the results of their testing. This approach prevents the measurement results from depending on the descriptive variety. In a previous work we realized a method and tool to test the state machine models, where instructors were required to manually create test cases. On the contrary, the method proposed herein provides a new function of fully and automatically generating test cases from the learner-created models. Thus, state machine models can be easily classified using various test cases that capture the generation-source model's behaviors. We validated the effectiveness of the method via application to 63 learner-created models that express the collaborative behaviors between two mobile robots.