How to create a new AML typology (alert transaction model)

1. Edit a Python script

First, you need to add a function to transaction_graph_generator.py to register a new AML typology.

1.1. Add process codes to generate alert transactions

Add process codes to add_alert_pattern function. For example, the fan_in pattern is defined as follows:

if typology_name == "fan_in":  # fan_in pattern (multiple accounts --> single (main) account)
            main_acct, main_bank_id = add_main_acct()
            num_neighbors = num_accounts - 1

            if is_external:
                sub_bank_candidates = [b for b, nbs in self.bank_to_accts.items()
                                       if b != main_bank_id and len(nbs) >= num_neighbors]
                if not sub_bank_candidates:
                    logger.warning("No banks with appropriate number of neighboring accounts found.")
                    return
                sub_bank_id = random.choice(sub_bank_candidates)
            else:
                sub_bank_id = main_bank_id
            sub_accts = random.sample(self.bank_to_accts[sub_bank_id], num_neighbors)
            for n in sub_accts:
                self.remove_typology_candidate(n)
                add_node(n, sub_bank_id)

            for orig in sub_accts:
                amount = init_amount
                date = random.randrange(start_date, end_date)
                add_edge(orig, main_acct, amount, date)

2. Create a new Java class

Next, create a Java class extends amlsim.model.aml.AMLTypology.