Visualisation and optimisation of learning classifier systems for multiple domain learning

Learning classifier system (LCSs) have the ability to solve many difficult benchmark problems, but they have to be applied individually to each separate problem. Moreover, the solutions produced, although accurate, are not compact such that important knowledge is obscured. Recently a multi-agent system has been introduced that enables multiple, different LCSs to address multiple different problems simultaneously, which reduces the need for human system set-up, recognises existing solutions and assigns a suitable LCS to a new problem. However, the LCSs do not collaborate to solve a problem in a compact or human observable manner. Hence the aim is to extract knowledge from problems by combining solutions from multiple LCSs in a compact manner that enables patterns in the data to be visualised. Results show the successful compaction of multiple solutions to a single, optimum solution, which shows important feature knowledge that would otherwise have been hidden.