| E-learning is becoming increasingly important in the field of education. Most existing systems do not focus on the learning styles of students. However, this trend is beginning to be reversed. Some systems developed in research laboratories focus on learner characteristics such as age and psychological type in order to appropriately assist users according to their needs.;In order to guarantee students a better interaction with the e-learning system from the beginning, we wanted to reduce the number of questions without compromising the quality of the psychological type recognition. The standardized questionnaire used is the Myers-Briggs Type Indicator (MBTI). The work presented in this thesis uses an approach of artificial neural networks and decision trees to determine which questions are less important for the effective recognition of the learner's psychological type and hence to reduce the size of the MBTI questionnaire.;Our experiments allowed a reduction of more than 33 % of MBTI questions. Such a reduction can also be beneficial for counsellors in schools who administer the MBTI questionnaire to students when advising them on their academic career. The results are general enough to be used by educators and e-learning software developers. Moreover, our experiment is a validation method of MBTI.;Key words: psychological types, MBTI, E-learning, adaptive systems, artificial neural networks, decision trees, machine learning.;Two methods exist to identify the learner's psychological type. In the first method, the psychological type of the learners is deduced from their interaction with the learning system. In the second method, the student completes any of the standard tests for the recognition of the psychological type [e.g., Honey & Mumford (1992) or the Myers-Briggs (1980)]. Next, a specialist interprets the results and assigns a corresponding psychological type so that the system is equipped to properly interact with the learner. |
