Forecasting high-quality patents from the perspective of technology convergence

Purpose High-quality patents are crucial for the advancement of science and technology. Therefore, the early identification of potential high-quality patents and recognizing the significance of promoting science and technology are priorities in this domain. This research aims to forecast high-quality patents from the perspective of technology convergence. Design/methodology/approach This study innovatively introduced technology convergence features, including convergence in the same field (CSF) and convergence in the different field (CDF). The research investigated six machine learning methods and determined the best-performing model for identifying potential high-quality patents. Findings Among the various machine learning models with 4 evaluation metrics (accuracy, recall, precision and F1), when introduced to a single feature (CDF or CSF), Random Forest is the best model in identifying potential high-quality patents. Nevertheless, AdaBoost demonstrates superior performance when combined with CSF and CDF. Originality/value This study innovatively introduced technology convergence features, including CSF and CDF.