Dedect relationship between 'a' and 'e' even 'f', 'g', 'h', etc by tracing through payments with similar amounts and identifying payment chains from the transaction data. It is a pattern graph search problem.
For example：
| id | accname | Event_Dt | Tx_Amt | Cntpty_Acct_Name |
|---|---|---|---|---|
| 5068 | a | 2020-01-01 | 20.0 | b |
| 5069 | a | 2020-01-02 | 300.0 | b |
| 5070 | a | 2020-01-03 | 180.0 | b |
| 5071 | b | 2020-01-03 | 40.0 | c |
| 5072 | b | 2020-01-03 | 500.0 | c |
| 5073 | b | 2020-01-04 | 10.0 | c |
| 5074 | b | 2020-01-03 | 150.0 | d |
| 5075 | c | 2020-01-04 | 50.0 | e |
| 5076 | d | 2020-01-04 | 150.0 | e |
You can run
spark-submit --files build_functions.cp37-win_amd64.pyd Aml.py in windows environment or spark-submit --files build_functions.cp37-x86_64-linux.so Aml.py on linux platform to get the payment chains seem like money laundering.
| id | chain_id | src | dst | amount | depth | accname | Event_Dt | Tx_Amt | Cntpty_Acct_Name |
|---|---|---|---|---|---|---|---|---|---|
| 5068 | 0 | a | e | 200.0 | 3 | a | 2020-01-01 | 20.0 | b |
| 5070 | 0 | a | e | 200.0 | 3 | a | 2020-01-03 | 180.0 | b |
| 5071 | 0 | a | e | 200.0 | 3 | b | 2020-01-03 | 40.0 | c |
| 5073 | 0 | a | e | 200.0 | 3 | b | 2020-01-04 | 10.0 | c |
| 5074 | 0 | a | e | 200.0 | 3 | b | 2020-01-03 | 150.0 | d |
| 5075 | 0 | a | e | 200.0 | 3 | c | 2020-01-04 | 50.0 | e |
| 5076 | 0 | a | e | 200.0 | 3 | d | 2020-01-04 | 150.0 | e |
Same 'chain_id' indicates these transactions belong to a complete chain. Field 'depth' shows the length of the chain. Enjoy it ! 🤗