QRIS Off-Hours and Periodicity Explorer

Overview

I built a Streamlit script to inspect transactions outside configured operating hours and flag merchants whose transactions occur at unusually regular intervals.

Objective

I used the experiment to ask:

1. Which transactions occur during a configured off-hours window?
2. Which merchants show repeated off-hours transactions with unusually consistent timing?

The output is an investigative flag, not a fraud determination.

What I Built

• CSV and Excel upload support.
• Configurable mapping for date, time, merchant, and amount fields.
• A configurable off-hours window that can cross midnight.
• Per-merchant transaction interval calculations.
• Filters based on minimum activity, median interval, and interval variability.
• Summary tables, visualizations, and downloadable flagged records.

Workflow

Uploaded transaction file
  -> column mapping and datetime parsing
  -> off-hours classification
  -> merchant-level interval calculation
  -> configurable periodicity rules
  -> flagged-merchant table and charts

Current Status

The current work is an exploratory Streamlit script. I have not established a formal test suite, benchmark dataset, calibrated anomaly model, or production-use result for it.

Limitations

• Regular intervals do not automatically indicate suspicious behavior.
• The rule thresholds are exploratory and require business validation.
• Merchant operating hours, timezone, seasonality, and transaction context are not fully modeled.
• Uploaded data may contain sensitive financial information and must be handled locally and securely.

Technology Stack

Python, Streamlit, pandas, NumPy, Matplotlib, Altair, Excel/CSV.

Publication Note

Public examples must use generated merchants and transactions. Real merchant identifiers, transaction times, amounts, source files, and investigation results remain private.