An evolutionary computing approach for mining most frequent subgraph by using size value evaluation

Graph databases have become one of the most attractive areas for researchers, especially in data mining field. Frequent subgraphs are one of the most actively researched in this area. Numerous algorithms have been developed to find the most frequent subgraphs, such as SUBDUE algorithm, which uses a computationally constrained beam. Consequently, it will lose the chance of producing suboptimal subgraphs. This approach aims to provide another method by employing an evolutionary computing algorithm for mining most frequent subgraphs using subgraph size value based on SUBDUE. The purpose is to take the advantage of a global search feature of evolutionary programming to discover different space solutions of subgraphs.We tested the performance of our algorithm on one artificial generated dataset, and two real datasets, and compared the performance of SUBDUE. Our results show that our algorithm could discover frequent subgraphs faster than SUBDUE, indicating the potential usefulness of our algorithm.