Improving the adaptability of multi-agent based E-learning systems

Francisco PINTO-SANTOS, Hector SÁNCHEZ SAN BLAS, Manuel SALGADO DE LA IGLESIA, Xuzeng MAO
2018 Advances in Distributed Computing and Artificial Intelligence Journal  
KEYWORD ABSTRACT multi-agent systems; e-learning; clustering E-Learning is a new learning approach that involves the use of electronic technologies to access education outside of a conventional classroom (Alonso Rincon,). The objective of E-Learning systems is to increase the students' learning skills by providing a customized experience to each system user (Rodrigues, 2013) . However, to accomplish this, it is necessary to monitor the continuous changes in the environment, mainly the students'
more » ... knowledge and skill acquisition. A multi-agent system architecture and a clustering algorithm are proposed for this purpose (as presented in (Rodrigues, 2014) This paper is an extension to the work of (Al-Tarabily, 2018) because it not only monitors changes in the student environment but also in the project environment, increasing the system's adaptability and accuracy. This paper presents a multi-agent-based e-learning system, as an extension to (Al-Tarabily, 2018). The architecture designed by (Al-Tarabily, 2018) is composed of five agents and uses the subtractive-DMAPSO algorithm (Kennedy, 1995) to cluster projects and students. Once the clusters are generated, the system executes the matching algorithms to optimally match the student clusters with the project clusters. The first agent used is PCA, it organizes the projects according to the level of skill that a student needs to accomplish them; then the SCA agent clusters the students by their skills; the third one is the SPMA, which maps the student groups formed in the SCA to appropriate projects, according to each group's average level of skill; the fourth is the SSMA which matches students with 'helper' students in order to complement their knowledge; finally, the DSCA, which surveys the environment searching for changes in students' skills and informs the SCA of those changes. The new proposal enhances the system's adaptability to changes in the environment by adding one more agent, the Dynamic Project Clustering Agent (DPCA), which does the same job as DSCA but instead of keeping track of the students' skills it evaluates the skill level required for the projects. Thanks to this agent it is possible to detect when a project is badly evaluated in terms of the skills required to carry it out, correct it and send these results to the PCA. By using these agents, it has been possible to map all the students and project correctly. Also, the 'helpers' perfectly complemented the students' skills. Finally, the DSCA and DPCA helped the system to adapt to changes dynamically. This paper is organized as follows: Section 2 reviews the current advances in the field of E-Learning. Section 3 proposes an architecture based on six agents inspired by (Al-Tarabily, 2018). Experimental results are outlined and discussed in 4. Finally, conclusions are presented in Section 5.
doi:10.14201/adcaij20187516 fatcat:sxqsoado2zcjxpk33bpf6fjdte