Multi-class Classification using BFS Crossover in Genetic Programming

Multi-class classification is a kind of classification task which involves processing an input object and then assigning this object to one of the more than two possible classes.Crossover operation is considered to be a primary genetic operator for modifying the program structures in Genetic Programming (GP). Genetic Programming is a random process, and it does not guarantee results. Randomness is a problem that occurs during crossover operation. During the standard breeding process in Genetic Programming, crossover operation produces offspring with less than half of the fitness of their parent. Thus, it reaches the state where performance stops increasing a certain point, which ultimately leads to unsatisfactory performance of the GP. In this paper, we are proposing a special type of crossover operation named as BFS (Best First Search) crossover to improve overall performance of crossover operation. The proposed method is categorized as best first search; this ensures that it finds the optimal and complete solutions. To demonstrate our approach, we have designed a multiclass classifier using GP and tested it on various benchmark datasets. The results attained show that by applying BFS crossover together with point mutation refined the performance of classifier.