leerish/aml-risk-intelligence-platform

Purpose and Entry Points

This is a batch AML analytics pipeline that converts PaySim-style financial transactions into SQL Server warehouse tables and Power BI investigation/reporting views. Primary execution assets are the analytical notebook at notebooks/AML-Analytics.ipynb, the database loader at src/data_to_sql.py, and SQL warehouse logic in sql/AML.sql.

Source Data and Analytics Processing

Python/Pandas performs dataset preparation and derives customer- and transaction-level fraud signals. The notebook is the likely interactive development and analytics entry point; src/data_to_sql.py operationalizes loading processed outputs into SQL Server. Raw/processed data directories described in the README are not present in this tree, so input-data location is an external/runtime dependency.

Customer Risk Profiling

Customer behavior is aggregated into the CustomerRisk analytical model: transaction volume, frequency, average amount, fraud history/rate, and derived risk category/score. This customer-level entity is the primary boundary between raw transaction events and compliance-oriented case prioritization. Relevant implementation locations: notebooks/AML-Analytics.ipynb, sql/AML.sql.

Alert and High-Risk Detection

The pipeline creates an investigator-facing AlertQueue from risk-ranked customers and identifies suspicious, high-value transaction records in HighRiskTransactions. These stages reduce the broader customer/transaction population into a reviewable operational scope. Core logic is represented in sql/AML.sql and the analytics notebook at notebooks/AML-Analytics.ipynb.

Transaction and Trend Intelligence

Transaction-level aggregation produces TransactionSummary, including fraud concentration by channel/type, while time-series rollups produce DailyTransactionMetrics for fraud and alert trend monitoring. These warehouse-ready models support exposure analysis, anomaly/spike visibility, and dashboard filtering. Paths: sql/AML.sql, notebooks/AML-Analytics.ipynb.

Warehouse Boundary and Connectivity

Microsoft SQL Server is the persistent analytics warehouse and integration boundary between computation and BI. Python connects through SQLAlchemy and PyODBC; src/data_to_sql.py is the load boundary and src/test_connection.py provides connection validation. SQL-side transformations/schema definitions are centralized in sql/AML.sql.

Reporting and Runtime Environment

Power BI Desktop consumes the warehouse models through the packaged report at dashboards/AML-Analtyics.pbix; dashboard screenshots belong under dashboards/screenshots. The runtime is Python 3.10-oriented Conda with Pandas, SQLAlchemy, and PyODBC dependencies defined by environment.yml and requirements.txt. External services/infrastructure are the PaySim dataset source, a reachable SQL Server instance, and Power BI Desktop.