Liquidity: Forecasting Pipelines & Dashboard

Overview

• End-to-end Kedro project that ingests dummy datasets (clients, EOD balances, remuneration, account types), processes and aggregates deposits, and produces a simple country-level forecast using interest rates.
• A Streamlit dashboard renders the processed aggregates and forecast results with a custom theme and consistent styling.

Highlights

• Modular, production-style Kedro pipelines with parameters.
• Forecasting at country level based on historical deposits and provided/observed interest rates (simple elasticity model).
• Streamlit dashboard with sidebar filters, sectioned charts, and consistent theme.
• Ruff-compliant code with docstrings and unit tests targeting 100% coverage for all node functions.

Project Layout

• src/liquidity/: Source package, pipelines, and registry.
• conf/base/: Kedro catalog and parameters.
• data/: Dummy data (raw) and pipeline outputs.
• app/streamlit_app.py: Dashboard app.
• .streamlit/config.toml: Theme config.
• tests/: Unit tests for nodes with full coverage.

Getting Started

1. Install dependencies (suggested):

   • Create a virtual env, then: pip install -e . or pip install -r requirements.txt (if you generate one).
   • Ensure Python >= 3.9.

2. Run the Kedro pipeline:

   • kedro run (from project root). Outputs will be written to data/08_reporting.

3. Launch the Streamlit app:

   • streamlit run app/streamlit_app.py

Parameters

• Edit conf/base/parameters.yml to tweak forecast horizon and default average interest rate (per-country overrides supported).

Notes

• The forecasting logic uses a simple linear elasticity model on log-deposits vs. interest rate to keep dependencies light and code transparent. It’s suitable for showcasing workflow and structure.
• Replace dummy CSVs in data/01_raw with real datasets to run the project with your data.