Annfelicty/Shilingi_AI

Smart, data-driven inflation forecasting for Kenya powered by machine learning, sleek visuals, and a little genius. # Commodity Price Forecasting Using Machine Learning ## Project Overview This project predicts the **next-month average prices** of key commodities in Kenya, using historical data, seasonality trends, and lag-based features. The goal is to provide **actionable forecasts** for stakeholders such as businesses, government agencies, and supply chain managers, helping them make informed procurement, pricing, and risk management decisions. The modeling approach focuses on: - High-quality **real-world price data** across multiple commodity types (food, energy, utilities, rent, etc.) - Awareness of **seasonality and market cycles** - **Unit normalization**, e.g., electricity in kWh - Explainable and interpretable machine learning models, primarily **Random Forest**, with metrics to track predictive accuracy ## Project Objectives 1. **Forecast Commodity Prices** Predict the next month’s average price for each commodity variant, enabling better planning and cost management. 2. **Understand Price Dynamics** Quantify short-term and seasonal trends, volatility, and persistence using lag, rolling averages, and change features. 3. **Support Decision-Making** Generate actionable insights for businesses, policymakers, and investors using interpretable ML models. 4. **Provide Robust and Scalable Tools** Implement per-commodity modeling pipelines that scale to hundreds of items with automated evaluation and reporting. ## Data The project uses historical commodity prices from 2012–2025. The dataset includes: | Column | Description | |--------|-------------| | `commodity_name` | Name of the commodity (e.g., Tomatoes, Electricity) | | `units_of_measure` | Unit for the price (e.g., kg, litre, 50 KWh) | | `current_average_price` | Monthly average market price | | `year`, `month`, `quarter` | Time-based features for seasonality capture | | `price_lag_1`, `price_lag_2`, `price_lag_3` | Lag features for last 1–3 months | | `price_rolling_mean_3`, `price_rolling_std_3` | Rollin …