On the Potential of Network-Based Features for Fraud Detection

By Catayoun Azarm, Erman Acar, Mickey van Zeelt

Rating

1402
Battle Count: 19

Relevance

7/10
While focused on fraud detection, the network-based approach could be adapted for detecting anomalies in trading patterns or assessing counterparty risk.

Implementation Complexity

6/10
Requires implementation of personalized PageRank and integration with existing fraud detection systems. Complexity increases with large-scale transaction networks.

Reproducibility

3/5
The paper provides details on the methodology and features used, but the dataset is proprietary and not publicly available.

About this paper

Methodology: Logistic Regression with Network Features. Problem types: Classification, Anomaly Detection.

The interactive Everscope explorer (charts, battles, favorites) loads below.