A comprehensive framework for patent value indicators: Machine learning-driven value prediction in renewable energy patents

Abstract

This study proposes a machine learning (ML) framework for predicting the value of renewable energy (RE) patents using a structured set of bibliometric and inventor-related indicators. Motivated by the strategic importance of clean technologies, a structured dataset of renewable energy patents is constructed using a technology-specific classification approach. In addition to widely used features, the study introduces and empirically tests three underexplored value indicators: first forward citation speed, lead inventor experience, and inventor team experience. Six ML algorithms; Naïve Bayes, Logistic Regression, Random Forest (RF), Extreme Gradient Boosting, Support Vector Machines, and Artificial Neural Networks (ANN), are implemented under five comparative strategy settings to evaluate classification performance. Results demonstrate that models trained with a comprehensive set of indicators significantly outperform those relying on limited features, with RF and ANN showing the strongest accuracy. SHapley Additive exPlanations are applied to interpret model behavior and quantify the influence of all indicators. While established citation-based measures retain importance, the newly introduced indicators provide consistent and meaningful contributions to model performance. The proposed framework advances patent valuation research by demonstrating how interpretable, data-driven methods can improve the prediction of valuable patents in policy-intensive technology sectors. These findings offer practical insights for technology managers, policy-makers, and innovation strategists seeking to enhance portfolio analysis, funding prioritization, and strategic decision-making in the context of green innovation.