I participated in a project with the Chongqing Bureau of Statistics, applying LLMs to economic data integration by unifying multi-source financial, textual, and sentiment features into a tensor representation, and built a five-dimensional enterprise data coupling network.
- Proposed a “Five-Dimensional Enterprise Coupling Network”, extending industrial risk analysis from single-factor evaluation to a multidimensional framework integrating financial, market, operational, technological, and socio-legal dimensions.
- Integrated multi-source heterogeneous data from CSMAR, WIND, SSE, and FinChina-SA; employed large language models (LLMs) to process unstructured financial reports and public opinion data, and developed tensor-based representation to unify multi-type data, addressing data heterogeneity and fusion limitations of conventional methods.