Patent-Level Analysis of Structural Innovation Differences Between the US and China Using Grey Wolf Optimizer and Explainable AI
Year: 2026
Type: conference-paper
Abstract: An analysis of national innovation systems is key to the evaluation of global technological competition. Patent information is a useful tool to study innovation capacity and technological development of countries. This study adopts a machine learning (ML) framework with large-scale patent data to explore the differences in structural innovation between the United States and China, covering the shortcomings of previous studies that failed to consider the influence of ML, explainable AI (XAI) and emerging trends. The dataset consists of 279,231 patent records with many firm, city, and national level indicators. Effective feature selection method was applied after pre-processing to find the most influential variables, which is a combination of Principal Component Analysis with Information Gain. Synthetic Minority Oversampling Technique was applied during model training in order to cope with the class imbalance problem. Several ML models have been tested such as Logistic Regression, K-Nearest Neighbour, Support Vector Classifier, Artificial Neural Network and Random Forest Classifier. A hybrid Grey Wolf Optimization–Random Forest model was also developed to optimize the hyperparameters of the model and improve its predictive performance. The experimental results show that the proposed optimized model perform best with accuracy of 98.66% and AUC of 0.99, after the preprocessing and oversampling process. XAI, specifically Local Interpretable Model-agnostic Explanations was used to make the models more interpretable in order to find the key factors affecting predictions. The results indicate that the level of national patent, economic indicators, and innovation intensity of a city are important in differentiating between the innovation structures of the two countries.
Language: English
Subfield: Management of Technology and Innovation
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7204475498