Transaction Fraud Report Part 4 — Smart Fraud Error Reporting

Four-part sequence: This is Part 4 of one progressive transaction-fraud interview exercise. It retains Parts 1–3 and adds prioritized error reporting.

1. Part 1: Verify Transaction Data Integrity.
2. Part 2: High-Risk Rule Validation.
3. Part 3: User Behavior Matching.
4. Part 4: Smart Fraud Error Reporting (you are here).

Implement the final cumulative behavior: check six CSV fields, apply two high-risk rules, compare behavior with a user baseline, and report prioritized errors.

Input

Each string in transactions is one unquoted CSV row with exactly six comma-separated fields, in this order: transaction_id,user_id,amount,payment_method,country,hour. Trim surrounding whitespace from every field. A field containing only whitespace is empty. Rows are processed in input order, and repeated transaction IDs remain separate rows.

Each row of baselines has six strings: [user_id,usual_country,start_hour,end_hour,min_amount,max_amount]. Every nonempty transaction row has exactly one matching user baseline. Hour and amount intervals are inclusive and do not wrap. Numeric fields on nonempty rows are valid decimal integers.

Review rules

1. If any transaction field is empty, assign only MISSING_FIELD. Do not run the later levels for that row.
2. Otherwise, assign AMOUNT_OUT_OF_RANGE when the amount is outside the inclusive global interval [normalMin, normalMax]. Assign BLOCKED_METHOD when the payment method appears in blockedMethods; membership is case-sensitive.
3. Compare the transaction with its user's baseline on three features: exact country, hour inside the baseline interval, and amount inside the baseline interval. If fewer than two of the three features match, assign BEHAVIOR_MISMATCH. This is the at-least-50% matching threshold for three features.

Report

Return a string table. Its first row is ["transaction_id","status","error_1","error_2"]. Append one four-cell row for every transaction, keeping the original transaction ID as its first cell, including an empty ID when that field is missing.

For each transaction, collect error codes in this priority order: MISSING_FIELD, AMOUNT_OUT_OF_RANGE, BLOCKED_METHOD, BEHAVIOR_MISMATCH. Put only the first two codes in the final two cells, using an empty string for any unused code cell. Use status OK when there is no error; otherwise use SUSPICIOUS. The four-cell rows form an aligned columnar report.