Optimisation of multiple classifier systems using genetic algorithms

摘要: We introduce a novel multiple classifier system with the ability of automatic self-configuration. The employs genetic algorithm to optimise configuration participating individual classifiers arranged in parallel structure. Our primary interest was study behaviour such an integrated system, first case increasingly complex tasks and secondly when additional information is made available form larger training data sets. fact that these cases often arise real world applications underline their special importance developing systems can address realistic problem domains. As example we tested proposed on character recognition task using one printed, handwritten, set. findings strongly suggest significant benefit be gained from integration algorithm-based optimisation process into both situations.