Genetic network programming with estimation of distribution algorithms and its application to association rule mining for traffic prediction

In this paper, a novel evolutionary paradigm combining Genetic Network Programming (GNP) and Estimation of Distribution Algorithms (EDAs) is proposed and used to find important association rules in time-related applications, especially in traffic prediction. GNP is one of the evolutionary optimization algorithms, which uses directed-graph structures. EDAs is a novel algorithm, where the new population of individuals is produced from a probabilistic distribution estimated from the selected individuals from the previous generation. This model replaces random crossover and mutation to generate offspring. Instead of generating the candidate association rules using conventional GNP, the proposed method can obtain a large number of important association rules more effectively. The purpose of this paper is to compare the proposed method with conventional GNP in traffic prediction systems in terms of the number of rules obtained.