Learning patterns in dynamic graphs with application to biological networks

摘要: We propose dynamic graph-based relational mining approach to learn structural patterns in graphs or networks as they change over time. There are a huge amount of data that can be represented graphs, and majority the have properties well properties. Most current approaches focus on only static but few address graphs. Our analyzes graph containing sequence discovers rules capture changes occur between pairs sequence. These represent rewrite first must go through isomorphic second graph. Then, our feeds into machine learning system learns general transformation describing types for class The discovered graph-rewriting show how time, repeated changes. We apply analysis dynamics biological with cell. A cell is not basic unit life, also an optimal system. This well-organized so it networks, which include various molecules relationships them. Moreover, their structure time express systems. In research, we understand biosystems evaluate results using coverage prediction metrics, compare those literature. important example, discovering known networks. Results learned accurately predict future networks. We two other data: synthetic Enron email data. several varied conditions, such noise, size density ratio. data, alternative approach.