Mining Temporal Subgraph Patterns in Heterogeneous Information Networks

With an increasing interest in social network applications, finding frequent social interactions can help us to do disease modeling, cultural and information transmission and behavioral ecology. We model the social interactions among objects and people by a temporal heterogeneous information network, where a node in the network represents an individual, and an edge between two nodes denotes the interaction between two individuals in a certain time interval. As time goes by, lots of temporal heterogenous information networks at different time unit can be collect. In this work, we aim to mine frequent temporal social interactions (call patterns) exist in numerous temporal heterogeneous information networks. We propose a novel algorithm, TSP-algorithm (Temporal Subgraph Patterns algorithm) to mine the patterns which contain temporal information and forms a connective subgraph. The proposed method recursively grows the patterns in a depth-first search manner. Since the TSP-algorithm only needs to scan the database once and does not generate unnecessary candidates, the experiment results show that the TSP-algorithm outperforms the modified Apriori on time-efficiency and memory usage in both synthetic and real datasets.