Extracting knowledge from complex graphs
Christos Faloutsos, CMU

PNC workshop
April 16, 2024

Abstract

How can ML/AI help with analyzing financial data? Such data are often in the form of graphs, like who-buys-what, and who-pays-whom; often with time-stamps, and often with labels (fraud/honest). In such cases, how can we find patterns, anomalies and discontinuities? How can we explain the results to a non-statistician? These are exactly the questions we focus on this tutorial. First, we will cover patterns we expect to see in real graphs, and time-tested graph-mining tools, like community detection and belief propagation. Then we will cover tools for mining time-evolving graphs, and specifically tensors and outlier-detection methods. Finally, we conclude with visualization tools for graphs, and specifically the recent, publicly available, CallMine tool

Bio

Christos Faloutsos is a Professor at Carnegie Mellon University. He is an ACM Fellow; he has published over 500 refereed articles, 17 book chapters and three monographs. He has received the SIGKDD Innovations Award (2010), and 31 ``best paper'' awards (including 8 ``test of time'' awards). His research interests include large-scale data mining with emphasis on graphs and time sequences; anomaly detection, tensors, and fractals.

Material - research articles and books

Material - Videos and code

Last updated: April 8, 2024, by Christos Faloutsos