上市公司人工智能概念炒作水平数据（2012-2024年）
Sou Hu Cai Jing·2025-12-22 00:34

Core Insights - The article discusses the measurement of artificial intelligence (AI) concept speculation among publicly listed companies, focusing on the authenticity of AI-related disclosures and opportunistic behaviors [1] - It highlights the importance of a quantitative approach to differentiate between substantial transformation and symbolic speculation in the AI sector, providing critical data support for market behavior analysis and regulatory improvements [1] Data Overview - The data is derived from three main modules: annual reports of A-share listed companies from 2010 to 2024, AI-related word frequency data, and financial metrics of listed companies from 2012 to 2024 [1] - The annual reports' "Management Discussion and Analysis (MD&A)" section serves as the primary source for strategic and technological disclosures [1] - AI-related word frequency data is generated through Python-based semantic analysis, utilizing authoritative reports as a basis for the dictionary [1] Data Selection and Cleaning - A strict standardization process is followed for data selection, starting with all A-share listed companies from 2010 to 2024, excluding ST companies, financial institutions, and those with missing annual report texts [2] - The Word2vec model is employed for semantic matching and frequency statistics of MD&A texts, resulting in an initial dataset [2] - A regression model is constructed to calculate residual indicators, incorporating company characteristics such as size, leverage, and return on assets, along with the logarithm of AI patent counts as key explanatory variables [2] Data Metrics - The dataset includes 10,230 records, ultimately yielding 7,872 valid observations that reflect the core measurement results of AI concept speculation levels among listed companies from 2012 to 2024 [2] - Key metrics include the number of AI technology patents, company size, leverage, return on assets, and Tobin's Q, which are essential for analyzing the relationship between AI disclosures and company performance [3][4]