AllanOnyonka-ltsm/kenya-economic-graph

# Kenya Economic Knowledge Graph A prototype knowledge-graph system for modeling economic interconnections in the Kenyan market. Tests whether graph-propagation (predicting downstream effects through entity relationships) outperforms simple baselines using historical event data. ## Project Overview This project builds a directed knowledge graph of Kenyan economic entities (companies, regulators, government agencies, and market participants) with curated relationships. It then simulates a "news expansion agent" that adds temporary event-driven edges, runs a backtest engine against historical events, and compares graph-propagation signals against baseline approaches. ### Key Entities (20 nodes) | Sector | Entities | |--------|----------| | **Telecom** | Safaricom, Airtel Kenya, Telkom Kenya | | **Banking** | KCB, Equity Bank, Absa Kenya, Cooperative Bank, NCBA, Standard Chartered (UK) | | **Energy** | KenGen, Kenya Power, Energy Regulatory Commission | | **Consumer Goods** | BAT Kenya, EABL, British American Tobacco (LSE) | | **Agriculture** | NCPB, Yara East Africa, Farmers (aggregated) | | **Fintech** | M-Pesa | | **Government** | Central Bank of Kenya | ### Relationship Types (21 edges) - supplier/customer (3), competitor (6), regulator/regulated (6), parent/subsidiary (2), partner (2), investor/investee (1), distributor (1) ## Directory Structure ``` kenya-economic-graph/ ├── src/ │ ├── graph_builder.py # Seed graph construction (20 nodes, 21 edges) │ ├── data_fetcher.py # Price data fetching (yfinance + synthetic) │ ├── expansion_agent.py # News expansion agent (7 simulated events) │ ├── backtest.py # Backtest engine (5 historical events) │ └── baselines.py # Baseline comparison (4 approaches) ├── nb/ │ └── exploration.ipynb # Jupyter notebook walkthrough ├── data/ │ ├── seed_graph.json # Seed graph (node-link format) │ ├── seed_graph.gexf # Seed graph (Gephi-compatible) │ ├── augmented_g …