MOQPSO: A new quantum-behaved particle swarm optimization for nearest neighborhood classification

As a kind of “lazy” learning method, nearest prototype has been successfully used in many pattern classification problems. In these methods, the main idea is that a collection of prototypes has to be found which precisely represents the input patterns and then the classifier assigns class labels based on the nearest prototype in the collection. In this paper, the quantum-behaved particle swarm optimization (QPSO) is first employed to find effective prototypes, and a new algorithm, called the multiple collapse-orthogonal crossover QPSO (MOQPSO) is presented in order to reduce the number of necessary prototypes, speed up the convergence and improve the classification result. In order to evaluate the performance of the new method, we compared the results of PSO, QPSO, and MOQPSO. Furthermore, it is also competitive with the classical SVM classifier and traditional nearest neighbor classifier. Some typical UCI datasets are used as test instances.