A multi-agent bio-inspired system to map learners with learning resources using clustering based personalization

Abstract

The work makes an examination of clustering methods in a multi-agent system which is fully decentralized. This has the goal of grouping agents that have similar data or objectives as in the case of traditional clustering. But, this adds to some more additional constraints wherein the agent will have to be in the same place as opposed to being collected within a centralized database. For doing this, it will connect to agents within a random network and will search for them in a peer-to-peer based fashion for the other agents that are similar. The primary aim here was to tackle the basic problem in clustering on the Internet scale thus creating methods where the agents may be grouped thus forming coalitions. For the purpose of investigating the decentralized approaches and their feasibility, the work presents the K-means clustering, the multi-agent Firefly Algorithm, (FA) and the Differential Evolution (DE). This is done for a reasonable number to times and will be surprisingly good. The results of the experiment prove that the multi-agent firefly clustering has better performance compared to that of a multi-agent K-Means clustering or a multi-agent DE clustering.