Mina-zuki/Nigeria-energy-performance

# Nigeria Energy Performance Intelligence ## Purpose Interactive analytical decision-support application based on `training data.xlsx`. The decision question is: **Where should management focus to improve crude-production resilience and gas utilisation?** It brings together crude production, crude losses, deferred production, gas utilisation, state contribution, contract-type output and 2023 crude-price benchmarks. The project supports comparison and prioritisation; it does not infer causes or equipment failures. ## Defensible ML conclusion The supplied data does **not** support a reliable supervised production, loss or deferral prediction model. The only integrated national dataset has five annual observations (2019-2023); other tables differ in grain and period. The app therefore deliberately provides no prediction, risk score, model accuracy claim or artificial model artefact. ## Run locally ```bash python build_data.py pip install -r requirements.txt streamlit run app.py ``` ## Included project outputs - `build_data.py`: reproducible extraction and clean CSV generation - `app.py`: Streamlit analytical decision-support app - `model/metadata.json`: ML feasibility decision and limitations - `data/`: cleaned analytical datasets created from the supplied workbook ## To make future ML defensible Add a consistently grained monthly/field-level panel with at least several years of observations, approved operational predictors, an outcome available at the time of prediction, and a held-out future period.