Difficulty estimation of exercises on tree-based search algorithms using neuro-fuzzy and neuro-symbolic approaches

Abstract

A central topic in the educational process concerns the engagement and involvement of students in educational activities that are tailored and adapted to their knowledge level. In order to provide exercises and learning activities of appropriate difficulty, their difficulty level should be accurately and consistently determined. In this work, we present a neuro-fuzzy and a neuro-symbolic approaches that are used to determine the difficulty level of exercises on tree-based search algorithms and we examine their performance. For the estimation of the difficulty level of the exercises, parameters like the number of the nodes of the tree, the number of children of each node, the maximum depth of the tree and the length of the solution path are taken into account. An extensive evaluation study was conducted on a wide range of exercises for blind and heuristic search algorithms. The performance of the approaches has been examined and compared against that of expert tutors. The results indicate quite promising performance and show that both approaches are reliable, efficient and confirm the quality of their exercise difficulty identification.