Learning Discriminating Functions Based on Genetic Programming for Classification

Knowledge discovery and data mining have become a hot research topic of late years. Classification is one of the most important problems in knowledge discovery. So many different classification algorithms have been developed for classifying data. In this paper, we present an effective scheme for classifying data with multi-category based on the technique of genetic programming. The proposed method presents a training strategy called adaptable incremental learning strategy and a well-defined fitness function to learn the set of discriminating functions from the given examples using genetic programming. For a k-class problem, after the k discriminating functions are generated, a Z-value measure is developed and used to resolve the problem of conflict. The experiments show that the proposed GP-based classification algorithm is efficient and has high accuracy. (