Big data analysis of policy coordination paths based on latent dirichlet allocation model and fuzzy-set qualitative comparative analysis method

The selection and coordinated application of government innovation policies are crucial for guiding the direction of enterprise innovation and unleashing their innovation potential. However, due to the lengthy, voluminous, complex, and unstructured nature of regional innovation policy texts, traditional policy classification methods often overlook the reality that these texts cover multiple policy topics, leading to lack of objectivity. In contrast, topic mining technology can handle large-scale textual data, overcoming challenges such as the abundance of policy content and difficulty in classification. Although topic models can partition numerous policy texts into topics, they cannot analyze the interplay among policy topics and the impact of policy topic coordination on enterprise innovation in detail. Therefore, we propose a big data analysis scheme for policy coordination paths based on the latent Dirichlet allocation (LDA) model and the fuzzy-set qualitative comparative analysis (fsQCA) method by combining topic models with qualitative comparative analysis. The LDA model was employed to derive the topic distribution of each document and the word distribution of each topic and enable automatic classification through algorithms, providing reliable and objective textual classification results. Subsequently, the fsQCA method was used to analyze the coordination paths and dynamic characteristics. Finally, experimental analysis was conducted using innovation policy text data from 31 provincial-level administrative regions in China from 2012 to 2021 as research samples. The results suggest that the proposed method effectively partitions innovation policy topics and analyzes the policy configuration, driving enterprise innovation in different regions.