Financial Services Cookbook

A collection of graph-native patterns for the banking and finance sector.

1. Anti-Money Laundering (AML)

Simple Money Circular Flow

gsqlterminal
Copy
// Detects if money moves from Account A back to Account A through 3-5 hops
CREATE QUERY detect_circular_transfer(VERTEX<Account> start_account, INT max_hops) FOR GRAPH FinanceGraph {
  OrAccum @visited = false;
  ListAccum<VERTEX<Account>> @path;

  Start = {start_account};
  
  // Traverse through transfers
  WHILE Start.size() > 0 AND max_hops > 0 DO
    Start = SELECT t FROM Start:s -(TRANSFER:e)-> Account:t
            ACCUM t.@path += s
            POST_ACCUM IF t == start_account THEN 
              PRINT "Circular flow detected", t.@path
            END;
    max_hops = max_hops - 1;
  END;
}

High Frequency Layering

gsqlterminal
Copy
// Detect accounts with more than 10 incoming transfers within 1 hour
CREATE QUERY layering_detection(DATETIME window_start) FOR GRAPH FinanceGraph {
  SumAccum<INT> @transfer_count;
  
  Accounts = {Account.*};
  
  Result = SELECT t FROM Accounts:s -(TRANSFER:e)-> Account:t
           WHERE e.timestamp BETWEEN window_start AND datetime_add(window_start, INTERVAL 1 HOUR)
           ACCUM t.@transfer_count += 1
           HAVING t.@transfer_count > 10;
           
  PRINT Result;
}

2. Fraud Detection

Synthetic Identity Check

gsqlterminal
Copy
// Find accounts sharing the same Phone or SSN but different Names
CREATE QUERY synthetic_identity() FOR GRAPH FinanceGraph {
  MapAccum<STRING, SetAccum<VERTEX<Account>>> @@shared_attrs;
  
  Users = {User.*};
  
  // Group users by SSN
  Result = SELECT s FROM Users:s
           ACCUM @@shared_attrs += (s.ssn -> s);
           
  // Filter for SSNs linked to multiple names
  FOREACH (ssn, user_set) IN @@shared_attrs DO
    IF user_set.size() > 1 THEN
      PRINT ssn, user_set;
    END;
  END;
}

3. Credit Risk

Path to Delinquent Entity

gsqlterminal
Copy
// Find exposure to a high-risk entity within 3 hops
CREATE QUERY risk_exposure(VERTEX<Entity> risk_source) FOR GRAPH FinanceGraph {
  SumAccum<INT> @dist = 0;
  OrAccum @is_exposed = false;

  Start = {risk_source};
  
  FOR i IN 1..3 DO
    Start = SELECT t FROM Start:s -(:e)-> Entity:t
            WHERE t.@is_exposed == false
            ACCUM t.@is_exposed = true, t.@dist = i;
  END;
  
  PRINT Start[Start.name, Start.@dist];
}

4. Merchant Analytics

Common Customer Analysis

gsqlterminal
Copy
// Merchants sharing the most customers
CREATE QUERY shared_customers(VERTEX<Merchant> m1, VERTEX<Merchant> m2) FOR GRAPH FinanceGraph {
  SumAccum<INT> @@common_count;
  
  Start = {m1};
  
  Customers = SELECT t FROM Start:s -(PURCHASED_AT:e)- Customer:t;
  
  Result = SELECT s FROM Customers:s -(PURCHASED_AT:e)- Merchant:t
           WHERE t == m2
           ACCUM @@common_count += 1;
           
  PRINT @@common_count;
}

5. Transaction Monitoring (Batch)

Daily Aggregate by Region

gsqlterminal
Copy
CREATE QUERY regional_daily_volume(DATETIME d) FOR GRAPH FinanceGraph {
  SumAccum<DOUBLE> @daily_sum;
  
  Regions = {Region.*};
  
  Result = SELECT s FROM Regions:s -(CONTAINS:e)- City:c -(HAS_ACCOUNT:e)- Account:a -(TRANSFER:t)- Account:a2
           WHERE t.timestamp BETWEEN d AND datetime_add(d, INTERVAL 1 DAY)
           ACCUM s.@daily_sum += t.amount;
           
  PRINT Result;
}

[!NOTE] These patterns can be scaled to billions of transactions by leveraging TigerGraph's massive parallel processing (MPP) capabilities.