The client needed to detect potential illegal money laundering activities by identifying shell companies through the analysis of business profile attributes and transaction patterns.

Developed an unsupervised ML model for shell company detection.

1. Feature Generation: Generated features involving business profile attributes (company age, revenue consistency) and transaction patterns (volume, frequency, directionality).

2. Feature Selection: Applied statistical techniques (correlation analysis, quasi-constant elimination, VIF, Fisher's method) to select the most relevant features.

3. Model Development: Used a combination of unsupervised machine learning algorithms to train the data and identify suspicious behavior for each entity in a cluster.

4. Pattern Generation: Predicted suspicious entities to generate networks and detect patterns based on network measures.Represented entities and interactions as nodes and edges in a graph for visualization, aiding the investigation.

Enhanced detection of shell companies and money laundering activities.

- The final model produced a 91% accuracy with a 94% precision rate and an 89% F1 score.

- The detection of suspicious entities improved the bank's ability to identify and prevent illicit activities, ensuring compliance and reducing financial crime risks.