Gene Transfer: A Novel Genetic Operator for Discovering Diverse-Frequent Patterns

Abstract

Genetic algorithm (GA) based on evolution principles has found its strong base in pattern set mining. GA has proved to generate more accurate results when compared to other formal methods available in the past years. In this paper, we present a new genetic operator called gene transfer within the GA framework. Genes are part of the genetic code that represents a partial solution to the problem. The proposed GA operator creates a pool of genes or partial solutions which are of good health and tries to replace genes of similar size in the population to improve the fitness of the individuals. This operation applied along with other traditional genetic operators like crossover and mutation results in speed up in convergence and finding individuals with better fitness function. On a set of standard benchmark dataset, we experimentally show that our new genetic operator improves the performance of a genetic algorithm consistently.