Frequent Pattern Mining in Continuous-Time Temporal Networks

Area of Science:

• Network Science
• Data Mining
• Computer Science

Background:

• Temporal networks are crucial in various disciplines.
• Frequent pattern mining is essential for network analysis.
• Existing methods often represent temporal networks as static sequences, leading to a computation-expressiveness trade-off.

Purpose of the Study:

• To propose a novel, lossless representation for temporal networks.
• To introduce constrained interval graphs (CIGs).
• To develop algorithms for mining frequent temporal patterns.

Main Methods:

• Developed a novel network representation preserving temporal aspects losslessly.
• Introduced and utilized constrained interval graphs (CIGs).
• Designed algorithms for mining complete sets of frequent temporal patterns, considering four isomorphism definitions.

Main Results:

• The proposed representation effectively captures temporal network dynamics.
• Algorithms successfully mined frequent temporal patterns from real-world datasets.
• Demonstrated the practicality and pattern discovery capabilities of the approach.

Conclusions:

• The novel representation and algorithms offer a significant advancement in temporal network analysis.
• This method overcomes limitations of static network representations.
• The approach is practical and effective for discovering unknown patterns in diverse temporal network settings.