Mini Lecture: Social Network Analysis for Fraud Detection

Veronique Van Varsar received a master's degree in Business Economics from G Len in Belgium with a thesis on mining data on Twitter, winning the best thesis award.

The speaker, a PhD researcher, discusses improving traditional fraud detection techniques with social network analytics by considering relationships between fraudsters and shared knowledge of fraud.

Exploration of challenges in fraud detection, including the need to integrate social knowledge, define fraud characteristics, and address issues like uncommonness, concealment, and time evolution of fraud.

The discussion on rebalancing data sets in fraud detection using Synthetic Minority Over-sampling Technique (SMOTE) to address the imbalance between fraudulent and legitimate cases.

An overview of utilizing social network analysis for fraud detection, extracting features, and rebalancing data to distinguish between fraudulent and non-fraudulent cases effectively.

Exploration of homophily concept in network analysis, focusing on similarities among individuals in terms of demographics or interests to identify potential fraudulent connections.

Explanation of spider constructions in fraud detection, where key companies are surrounded by side companies engaged in fraudulent activities, illustrating potential fraud patterns.

Final remarks on the importance of social network effects in fraud detection, emphasizing the need to focus on neighborhood features, fraud propagation, and anticipatory measures for effective fraud detection.