Predictive factors of citizen investment in rooftop solar panels: Evidence from a social innovation project in Iran using machine learning

Ali Asghar Sadabadi, Zohreh Rahimirad,
Predictive factors of citizen investment in rooftop solar panels: Evidence from a social innovation project in Iran using machine learning,
Journal of Cleaner Production,
Volume 529,
2025,
146813,
ISSN 0959-6526,
https://doi.org/10.1016/j.jclepro.2025.146813.
(https://www.sciencedirect.com/science/article/pii/S0959652625021638)
Abstract: In most oil-rich countries, despite their high solar potential, this capacity has been underutilized, and citizen participation in solar projects has remained limited. Social innovation (SI), particularly in the field of participatory financing, can create innovative, community-based models to mobilize the financial resources needed for renewable energy (RE) projects and reduce investment barriers. Therefore, accurately identifying the factors influencing citizens’ investment in this sector is essential for designing effective policies, financial instruments, and implementation models. We examined a national-scale SI program in Iran and analyzed data from 2077 participants in the installation of residential rooftop solar systems. We applied four machine learning algorithms—Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN)—to predict investment behavior. The RF model achieved the highest test accuracy (77 %) and revealed that connection to the national electricity grid, access to loans, government incentives, and understanding of benefits were the most important predictors. Public attitude towards solar energy, peer pressure, gender, type of residence, and homeownership status were also significant, whereas household income, education, and high solar potential had comparatively lower impact. The findings suggest that in oil-rich economies, removing financial barriers should be complemented by social interventions. From a policy perspective, these results underscore the need to design segmented programs based on demographic and locational characteristics, utilizing tools such as PAYG financing, crowdfunding models, demonstration projects in rural and underserved areas, and leveraging the influence of social networks to enhance participation. By linking data-driven analysis with SI, this study provides a practical framework for policymakers and RE stakeholders to accelerate and broaden the transition to clean energy.
Keywords: Social innovation; Renewable energy investment; Machine learning prediction; Residential rooftop solar panels; Oil-rich countries