solar-pv-healthcare-financial-viability-colombia

This dataset supports the article "Predictive Machine Learning Models for the Financial Viability of Sustainable Projects: A Case Study in Healthcare Facilities in Colombia." It provides energy consumption data, a financial cash-flow model, and dual Machine Learning results (binary classification of financial viability and continuous regression of Net Present Value) for 56 healthcare facilities in Barranquilla, Colombia. Consumption data originate from 2018 monthly billing records for 56 facilities (hospitals, CAMINO centers, and PASO primary-care centers). Electricity tariffs were updated to the values published by the regional utility Air-e for the 2025-2026 period (907, 943, 796, and 796 COP/kWh across the verified tariff checkpoints). For each facility, a 25-year discounted cash-flow model estimates CAPEX, annual O&M cost, avoided energy cost, Net Present Value (NPV), and Internal Rate of Return (IRR) under self-consumption solar PV. The repository includes: (1) the processed facility-level dataset with all technical-financial variables and the two ML target variables (Viabilidad_Financiera, VAN_COP); (2) PCA-reduced training and test sets used for the classifier and regressor; (3) classifier outputs (classification report, PCA feature importances); (4) regressor outputs (R²/MAE/RMSE metrics, PCA feature importances); (5) six figures covering correlation analysis, PCA explained variance, and exploratory consumption visualizations; and (6) the full manuscript and the MDPI citation style file. Note on traceability: the included classifier/regressor metrics correspond to a small validation split (45 train / 12 test) delivered with this data release. They differ from the final metrics reported in the article body (AUC = 0.993, F1 "not viable" class = 0.744, regression R² = 0.743), which were obtained using 5-fold GroupKFold cross-validation grouped by facility over the full 5,400-scenario Monte Carlo simulation. This dataset is therefore best understood as the facility-level processed data and a representative ML run, not a byte-for-byte reproduction of the article's headline metrics, unless a future version adds the complete Monte Carlo scenario file. Keywords: machine learning; solar photovoltaic; financial viability; Monte Carlo simulation; net present value; Gradient Boosting; XGBoost; healthcare facilities; energy poverty; Colombia; Barranquilla; Air-e tariff. License: CC BY 4.0. This dataset does not constitute financial, technical, or investment advice; it is intended for academic and research purposes only.