Deciphering Pico-Solar Product Adoption : A Random Forest Approach

What the paper says

Purpose : This paper examined the adoption of pico-solar products in rural India through the lens of diffusion of innovation (DOI). This study aimed to explore the effect of relative advantage, compatibility, complexity, and affordability on the adoption of pico-solar products and identify key determinants boosting the adoption. Methodology : Primary data were collected from 568 users and potential users of pico-solar products in the rural areas of the Pune district of India through a structured questionnaire. In order to better capture the non-linear relationships among the variables, a machine learning (ML) method – Random Forest was employed. The model performance was measured using the metrics MAE, RMSE, R2, and cross-validated R2. SHAP analysis, feature importance, and partial dependence plots (PDPs) were used for the analyses. Results : SHAP analysis and feature importance evaluation showed that relative advantage is most important in driving the adoption, followed by affordability, compatibility, and complexity. PDPs verified that the individual’s relative advantage perception, affordability, and compatibility enhanced adoption, and increased complexity decreased adoption likelihood. Practical Implications : The investigation showed that solar choice is not only determined by financial factors, but it also considers easy product usability, compatibility with rural lifestyles, and advantages over conventional energy. These implications would guide policymakers, solar enterprises, agencies, and firms regarding the development of marketing strategies focusing on financial incentives, easy product designs, and consumer training programs to boost pico-solar product adoption. Value : This work is a pioneering study to combine DOI and random forest modeling in the field of pico-solar products adoption.