Development of the Improved Exercise Generation Metaheuristic Algorithm EGAL+ for End Users

Abstract

Exercise generation is a subject worthy of investigation. In our previous papers, we presented a novel multi-objective harmony search metaheuristic algorithm called EGAL designed to address a widely recognised problem: generating diverse exercises to measure students’ knowledge on various topics. We demonstrated how to generate more subsets of predefined tasks (i.e. exercises) to measure students’ knowledge in such a way that the quality of these subsets should be good enough according to a predefined quality matrix (i.e. they should cover as many areas of the course as possible). Exercises should be diverse (based on the diversity measure, which is inserted into the multi-objective fitness function) and although the difficulty of these tasks can vary, the difficulty values of their subsets should be equal to generate fair exercises with an equal level of difficulty. The optimisation algorithm has been developed for professionals skilled in optimisation theory. The applications were not end-user friendly, as a high level of special skills and knowledge were needed for its implementation. According to our hypothesis, the EGAL algorithm can be modified for end users without a computer science and optimisation background with certain restrictions and improvements. An improved metaheuristic algorithm (EGAL+) was created and presented in this study. Our hypothesis for this improved algorithm was confirmed by running it on a large number of samples.