Transaction Risk Engine

Build an in-memory risk engine in core Java that marks each transaction as SAFE or RISKY using that account’s recent history. The current implementation marks every transaction as RISKY. Your job is to update RiskEngineImpl to apply the real rules so that all tests pass.

---

Rules

1. HIGH_FREQUENCY — RISKY if the account has more than 5 transactions in the last 2 minutes (exactly 5 is SAFE).
2. AMOUNT_ANOMALY — RISKY if the amount is greater than 3× the average of the prior last 10 transactions for that account. The current transaction is not included in that average. If there are fewer than 10 prior transactions, use the average of all prior transactions; if there are no prior transactions, do not trigger this rule. Exactly 3× is SAFE.
3. MERCHANT_DIVERSITY — RISKY if the account has more than 3 unique merchants in the last 5 minutes (exactly 3 is SAFE).

Reasons: HIGH_FREQUENCY, AMOUNT_ANOMALY, MERCHANT_DIVERSITY.

When more than one rule triggers: Include all triggered reasons in the result. A transaction can be RISKY for multiple reasons (e.g. high frequency and amount anomaly); the output must list every reason that applied. Tests verify that multi-reason outcomes are correct.

The first transaction for an account has no history; evaluate it as SAFE (no rules triggered).

---

Sample data: sample-transactions.csv

• 31 transactions, accounts A001–A007. Use this file to run the app and see output in the terminal.
• Format: transactionId,accountId,amount,timestamp,merchantId (first line is header). Timestamp is in milliseconds since epoch (epoch ms).
• Intended outcomes: A001/A005 normal → SAFE; A002 → T009, T019 RISKY (high frequency); A003 → T013 RISKY (amount anomaly); A004 → T017 RISKY (merchant diversity); A006 → T025 RISKY (amount anomaly); A007 → T031 RISKY (multiple reasons: HIGH_FREQUENCY, AMOUNT_ANOMALY, MERCHANT_DIVERSITY).

Intended Output

Transaction T001: SAFE Transaction T002: SAFE Transaction T003: SAFE Transaction T004: SAFE Transaction T005: SAFE Transaction T006: SAFE Transaction T007: SAFE Transaction T008: SAFE Transaction T009: RISKY [HIGH_FREQUENCY] Transaction T010: SAFE Transaction T011: SAFE Transaction T012: SAFE Transaction T013: RISKY [AMOUNT_ANOMALY] Transaction T014: SAFE Transaction T015: SAFE Transaction T016: SAFE Transaction T017: RISKY [MERCHANT_DIVERSITY] Transaction T018: SAFE Transaction T019: RISKY [HIGH_FREQUENCY] Transaction T020: SAFE Transaction T021: SAFE Transaction T022: SAFE Transaction T023: SAFE Transaction T024: SAFE Transaction T025: RISKY [AMOUNT_ANOMALY] Transaction T026: SAFE Transaction T027: SAFE Transaction T028: SAFE Transaction T029: SAFE Transaction T030: SAFE Transaction T031: RISKY [HIGH_FREQUENCY, AMOUNT_ANOMALY, MERCHANT_DIVERSITY]

Terminal output and commands

Build

mvn clean install -DskipTests

Run the app (see results in terminal)

mvn clean package && java -jar target/risk-engine-1.0.jar < sample-transactions.csv

Output format (one line per transaction)

• SAFE: Transaction <id>: SAFE
• RISKY (one reason): Transaction <id>: RISKY [HIGH_FREQUENCY]
• RISKY (multiple reasons): Transaction <id>: RISKY [HIGH_FREQUENCY, AMOUNT_ANOMALY] — list all reasons that triggered, in any order.

Example

Transaction T001: SAFE
Transaction T009: RISKY [HIGH_FREQUENCY]
Transaction T013: RISKY [AMOUNT_ANOMALY]

Run tests

mvn clean test

---

Performance and optimization

• Evaluate each transaction using only that account’s history; avoid full rescans.
• Use data structures that support time-window queries and incremental averages (e.g. ordered by timestamp, bounded windows).
• Tests expect: 10k transactions in < 2s; single-account 10k in < 3s; scaling better than linear in history size. Old data should be removable to keep memory bounded.

Success: App run on sample-transactions.csv shows expected SAFE/RISKY lines, and mvn clean test passes.