Flag Suspicious Transactions - Coding Interview

Question

## Problem

Given a list of transactions `(account_id, amount, timestamp)` sorted by timestamp, flag any transaction as suspicious if:
1. The amount exceeds 3x the account's rolling 7-day average (excluding the current transaction), OR
2. Three or more transactions occur from the same account within any 10-minute window.

Return the list of suspicious transaction indices (0-based, in original order).

```python
def flag_transactions(txns: list[tuple[str, float, int]]) -> list[int]:
    # txns[i] = (account_id, amount, timestamp_seconds)
    ...
```

**Example:**
```
txns = [
  ("acc1", 100, 0),
  ("acc1", 110, 86400),
  ("acc1", 800, 172800),   # -> flagged: 800 > 3*105
  ("acc2", 50,  0),
  ("acc2", 50,  300),
  ("acc2", 50,  500),      # -> flagged: 3 txns in 500s window
]
Output: [2, 5]
```

## Follow-ups
- Which rule is more expensive to compute? How would you optimize for a stream of real-time transactions?
- How do you handle accounts with fewer than 3 historical transactions for the rolling average?
- Extend rule 1 to flag only if the deviation exceeds 3 sigma (standard deviation-based threshold).

Flag Suspicious Transactions - Coding Interview

Question Details

Full Details

Topics

More from Bloomberg