Comparing hybrid versus single strategy intelligent systems in signal pattern classification

We define a hybrid intelligent pattern classifier for sensor systems such as an electronic nose. In our architecture, genetic algorithms evolve pattern templates, fuzzy logic identifies class boundaries, and neural networks refine these boundaries. The system exhibits superior performance and reasonable cost in handling large sets of patterns and noisy data. In this paper we use synthetic data to demonstrate the superior power of our hybrid system over single strategy systems. We compare performance, the classification cost and the cost of building the hybrid intelligent classifier to fuzzy clustering, probabilistic neural network and genetic algorithm classification systems