Computer Engineering and Informatics

The aim of this paper is to describe an application of evolutionary computation techniques dealing with the problem of constructing a student model for an intelligent tutoring system. Most traditional systems usually represent student knowledge and preferences by using a list of appropriate characteristics. The problems that arise in such cases are of two kinds: a) the list is comparatively small so that the tutoring system will be able to compute the student profile in a reasonable time. As a result, the list lacks many significant student features and the resulting student profile is not the desirable one. b) The list is too large in order to include all the possible useful behavioral and cognitive student features. In this case, the search for useful information in a large list of characteristics is a difficult and time-consuming task that the tutoring system has to perform.
According to the previously mentioned, there is a need for optimisation of the student model, so that it fits to as many student preferences and knowledge characteristics as possible, in a rather short time. In order to overcome these problems and search the information based on the optimal student’s features, we propose a novel algorithm based on evolutionary computation to extract an optimal student profile. This method originates from the simulation of the human immune system for finding antibodies (binary strings) that have the ability to detect (fit to) a large stack of intruders (strings). The proposed algorithm is trained with a collection of N different student profiles that are partitioned in k classes. The output of the algorithm is an ordered list of k student profiles, each one fitting to the features of one of the k classes. The algorithm has been used and tested as a part of a general tutoring system and we have achieved quite promising results.