Machine learning-based framework for assessing the financial viability of decentralized rainwater harvesting systems

Shiguang Chen, Qi Li, Hongwei Sun, Yongning Wang, Xuebin Chen,
Machine learning-based framework for assessing the financial viability of decentralized rainwater harvesting systems,
Results in Engineering,
Volume 28,
2025,
107215,
ISSN 2590-1230,
https://doi.org/10.1016/j.rineng.2025.107215.
(https://www.sciencedirect.com/science/article/pii/S2590123025032700)
Abstract: Urban rainwater harvesting (RWH) systems are essential for sustainable water management, but their adoption is limited by the challenges in assessing their economic feasibility. Traditional methods often fail to account for the nonlinear and multivariable factors affecting RWH systems. This study develops a machine learning-based framework to overcome these limitations and predict the economic viability of decentralized rooftop RWH systems. Using data from 1,841 RWH systems across 20+ countries and 150+cities, four machine learning algorithms—Random Forest (RF), AdaBoost, Light Gradient Boosting Machine (LGBM), and Multi-Layer Perceptron (MLP)—were evaluated. The RF model demonstrated optimal performance for Benefit-Cost Ratio (BCR) prediction (R²=0.906, MAE=0.096, RMSE=0.209), outperforming the Net Present Value (NPV) (R²=0.939, MAE=49,249, RMSE=512,453) and Payback Period (PBP) models (R²=0.883, MAE=13.53, RMSE=32.74). SHAP analysis identified water tariffs (20.8%) and rainfall variation (20.5%) as key BCR predictors; the catchment area (43.3%) and total built area (36.3%) dominated NPV forecasts, while project lifetime (19.7%) was the primary factor for PBP. Additionally, a user-friendly web application was developed using Streamlit, enabling users to input regional and building-specific parameters to assess the economic feasibility of RWH systems. This framework offers a robust tool for engineers, property owners, and policymakers to make informed decisions, promoting the global adoption of sustainable water management practices.
Keywords: Rainwater harvesting; Financial viability; Machine learning; (RFA)-Shapley; Explainability; Streamlit application