motheomoshageng/indabax-botswana-2026

# IndabaX Botswana 2026 — VenturePulse Hackathon ## Food Price Inflation Forecasting for Botswana ### Overview This repository contains my submission for the Deep Learning IndabaX Botswana 2026 Hackathon. The goal was to forecast Botswana's monthly food price inflation (FAO Item Code 23014) for January through December 2024 using five separate datasets at different granularities. ### Files in This Repository | File | Description | |------|-------------| | `predictions_2024.csv` | Final 12-month forecast (ARIMA model) | | `hackathon_notebook.ipynb` | Full pipeline — data loading, feature engineering, both models | | `feature_engineering_report.pdf` | Deliverable 1.1b — Justification of all features, lags, and merge strategy | | `model_comparison_report.pdf` | Deliverable 1.1c — ARIMA vs LSTM comparison and analysis | | `hcp_linkage_memo.pdf` | Deliverable 1.2a — Policy memo on regional food price dynamics | | `hcp_visualisations.png` | Deliverable 1.2b — Regional comparison and forecast charts | ### How to Run #### Requirements Install the required packages: ```bash pip install pandas numpy matplotlib xgboost scikit-learn statsmodels torch Step 1: Data Preparation Place the five CSV files in the same directory as the notebook: 01_baltic_dry_index_daily.csv 02_brent_crude_monthly.csv 03_botswana_policy_rate.csv 04_fao_botswana_prices.csv 05_human_capital_project.csv Step 2: Run the Notebook Open hackathon_notebook.ipynb in Jupyter and run all cells in order: Cells 1-4: Load and explore all five datasets Cells 5-7: Feature engineering — extract 10 monthly features from daily BDI, create lag features, merge all datasets Cell 8: ARIMA classical baseline — train and generate 2024 forecast Cell 9: LSTM deep learning model — train and generate 2024 forecast Cell 10: Generate visualisations and save prediction CSV Step 3: Output predictions_2024.csv — 12 rows, columns: year_month, forecast hcp_visualisations.png — Regional comparison and forecast charts Mo …