Predict which invoices will be paid late so you can act early
Advanced24h2–3
PythonScikit-learnFastAPI
17/20Transaction Anomaly Detector
Spot unusual transactions with explainable reasons.
Advanced24h2–3 people17/20 fit
Build Transaction Anomaly Detector for fintech teams and power users in ~24h with 2–3 people. Stack: Python + Scikit-learn + FastAPI.
Why it wins: Machine learning that judges can verify.
Build flow
01 · Problem
Fraud and errors hide among thousands of normal transactions.
02 · Solution
Use an isolation forest and per-user baselines to flag anomalies, then explain each flag in plain language for review.
03 · Build
- 1Generate sample data
- 2Train detector
- 3Explain flags
- 4Build review UI
04 · Demo
Inject a fake unusual charge and show it flagged with a reason.
MVP checklist
- Per-user baselines
- Anomaly scores
- Explanations
- Review queue
Tech stack
| Tech | Role | |
|---|---|---|
| Python | Backend / scripting | Backend |
| Scikit-learn | Tooling | Tool |
| FastAPI | External data | API |
Related ideas
More FintechPaste a suspicious payment message and get a scam risk verdict
Intermediate14h1–2LLMNext.jsClassifier17/20Fintech
Invoice Fraud Flagger
Catches duplicate invoices and changed bank details before payment
Advanced36h2–4PythonOCRLLM16/20