Frequent Subgraph Mining by Giraph Distributed System

Abstract

To overcome the challenges for managing the rapid growth of social graphs, massive Distributed Graph Mining Systems are developed, such as Pregel, GiraphHama, GraphLab, PowerLab, etc. The common approach to all systems is to divide the entire Graph Dataset into smaller divisions and use it as “think like a vertex”, the programing model is to hold up a continual graph calculation. In this paper, we use the Optimized Frequent Subgraph Mining algorithm in the Giraph framework model and make a comparative study with existing different Distributed Systems. To enhance the flexibility and performance of the novel method, we carry out different optimization techniques associating it with updating different run time limits. We also investigate how the performance could be improved by Giraph Distribution System, which plays a vital role in social graphs such as LinkedIn, Twitter, Facebook, etc. The graph input, output, cluster set up and hardware configuration play vital roles in optimizing the performance of our proposed algorithm.