Data-Driven Identification of Industrial Clusters: A Patent Analysis Approach

Abstract

Accurate identification of industrial clusters (IIC) serves as a reference for regional economic policymaking and enterprise development decision-making. Although data-driven methods have been extensively used in previous studies to support objective and effective work, both the data sources and research algorithms have significant shortcomings for IIC. To address these challenges, this article proposes a novel research framework that integrates patent mining and machine learning. Patents, with their quantifiable knowledge attributes and accessibility from public databases, are particularly suited for macrolevel analysis of innovation activities, providing robust support for identifying and analyzing clusters on a national scale, especially knowledge-intensive ones. This article introduces an improved density-based parameter adaptive algorithm designed to effectively carry out IIC based on the geographical location of patent applicants. Based on spatial cluster types defined by Markusen (1996), target clusters are classified using patent analysis. Four quantitative indexes–scale, output, efficiency, and quantity–are proposed to evaluate clusters based on their spatial structure and industrial organization. The practical application is demonstrated through a case study of China's flexible electronics industry. In addition, the Silhouette Coefficient index is employed to compare the effectiveness of the proposed algorithm against other methods. This article advances the theory of IIC, and provides foundation for scholars, calling for empirical research on industrial clusters from the perspective of individual enterprises. It also provides practical guidance for enterprises and policymakers on the application of IIC.