Please enable JavaScript.

Coggle requires JavaScript to display documents.

Data Mining Techniques for AML (Clustering (An outlier detection model…

- - - - Data:
        640 accounts and 120,986 original transaction records
        Features:
        transaction time, account number, transaction direction, and transaction amount
      - Anomalies:
        64 accounts and 1,940 transaction records reported from the bank and identified by the supervision center
      - Model:
        transactions modeled as temporal sequences
        use probabilistic model to find high-risk points, then add its former and following transaction to form high-risk sequence
      - 2008
  - - - Model:
        assume the money laundering transactions can form a graph pattern: a single sender -> multiple routes through various intermediaries -> final receiver
        Mine transactions to find such subgraphs
      - Data:
        synthetically generated by simulating a "mini-economy"
      - Anomalies:
        synthetic transactions that are sent from the sender to one intermediary and then to the receiver
      - 2011
    - - 2012
      - Model:
        assume the money laundering pattern is:
        money is gathered into a single account and commission is received by entities involved in the money laundering process
        Cluster transactions and find such money laundering pattern
      - Data:
        synthetically generated