Refactor and clean up codebase: remove unused scripts, enhance behavi…

[Samarthuday]

…oral features, and update model training pipeline

• Deleted serve_dashboard.py, check_status.py, simple_ingestion.py, start_system.py, and test_ingestion.py as they are no longer needed.
• Updated behavioral_features.py to include Payment_currency and Received_currency in partitioning for more accurate feature calculations.
• Modified baseline.py to include StandardScaler in the logistic regression pipeline for better model performance.
• Adjusted train.py to reflect changes in feature names and removed deprecated features.
• Added tests in test_leakage.py to ensure no leakage occurs and that behavioral features are correctly calculated based on currency.

[Copilot]

🟡 Changes recommended

There are confirmed inconsistencies/bugs (currency-aware refactor is incomplete for pair_transaction_count, and README Python version conflicts with pyproject.toml/CI) that should be resolved before approval.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR refactors the repository from a legacy “real-time monitoring demo” into a research-oriented modeling pipeline, focusing on leakage-safe feature engineering, improved baseline modeling, and reproducibility/CI hygiene.

Changes:

• Removed legacy dashboard/API/ingestion launcher scripts and old notebooks; updated README to the new research/pipeline framing.
• Updated behavioral feature generation to be more currency-aware and adjusted model feature contracts accordingly.
• Improved the logistic baseline (scaling) and strengthened metadata/reproducibility (package versions, training windows, probability quantiles), plus added CI and leakage-related tests.

File summaries

File	Description
tests/test_leakage.py	Adds tests for leakage-free behavioral features, currency awareness, and chronological split boundaries.
start_system.py	Deletes top-level legacy launcher script.
src/utils/test_ingestion.py	Deletes legacy ingestion test utility.
src/utils/start_system.py	Deletes legacy “start API + dashboard” orchestration script.
src/utils/simple_ingestion.py	Deletes demo transaction stream simulator.
src/utils/check_status.py	Deletes legacy status checker.
src/models/train.py	Updates model feature lists and retains chronological split helper.
src/models/baseline.py	Wraps SGD logistic baseline in a pipeline with StandardScaler.
src/features/behavioral_features.py	Updates rolling windows to partition by currency and refactors lifetime/network-related features + HHI logic.
src/dashboard/serve_dashboard.py	Deletes dashboard HTTP server.
src/dashboard/real_time_dashboard.html	Deletes legacy dashboard UI.
src/api/simple_api_server.py	Deletes legacy simple API server.
scripts/train_model.py	Improves artifact metadata (quantiles, time windows, prevalence, params, package versions).
README.md	Rewrites documentation for the research/pipeline focus and new API entrypoints.
pyproject.toml	Adds packaging/test/lint configuration and declares Python version requirement.
notebooks/03_backtesting.ipynb	Deletes legacy notebook.
notebooks/02_model_research.ipynb	Deletes legacy notebook.
notebooks/01_eda.ipynb	Deletes legacy notebook.
docs/assets/.gitkeep	Adds placeholder for documentation assets directory.
.github/workflows/ci.yml	Adds CI workflow to run tests on Python 3.11/3.12.

Review details

Suppressed comments (1)

src/features/behavioral_features.py:109

• pair_transaction_count is still partitioned only by (Sender_account, Receiver_account) while other rolling behavioral windows were updated to be currency-aware. This can mix interactions across currencies (e.g., USD and EUR transfers between the same pair) and makes the feature contract inconsistent with the currency-aware intent of this refactor.

            COUNT(*) OVER (
                PARTITION BY Sender_account, Receiver_account
                ORDER BY timestamp
                RANGE BETWEEN UNBOUNDED PRECEDING
                AND INTERVAL '1 microsecond' PRECEDING
            ) AS pair_transaction_count

• Files reviewed: 19/20 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.