我国区域人工智能产业政策的量化评价——基于主题挖掘与PMC-AE指数模型

Quantitative Evaluation of Regional Artificial Intelligence Industrial Policies in China—Based on Topic Ming and PMC-AE Index Model

摘要: 我国人工智能政策体系正处于从“量的积累”向“质的提升”转变的关键阶段，对其进行科学、系统的量化评估，有助于理解现有政策的效力与局限，可为政策的优化与区域协同提供关键依据。本文利用LDA主题模型提取政策主题，并构建评价体系，通过自编码器改进的PMC-AE指数模型，量化评价我国六大经济圈2017~2025年的人工智能产业政策，旨在揭示政策内在结构、区域差异与改进路径。研究发现，现有政策主题分布广泛，整体政策水平处于“一般”至“良好”的过渡阶段，但存在区域间政策供给、质量不平衡现象。由此，对先发型地区和追赶型地区分别提出了政策制定建议，并给出了提升政策整体效能的举措，有助于人工智能产业的高质量发展。

Abstract: China’s artificial intelligence policy system is at a critical stage of transformation from “quantitative accumulation” to “qualitative improvement”. Conducting a scientific and systematic quantitative assessment of it can help understand the effectiveness and limitations of the existing policies and provide a key basis for policy optimization and regional coordination. This paper uses the LDA topic model to extract policy themes and construct an evaluation system. Through the PMC-AE index model improved by the autoencoder, the artificial intelligence industry policies of China’s six major economic circles from 2017 to 2025 are quantitatively evaluated, aiming to reveal the internal structure, regional differences and improvement paths of the policies. Research findings show that the current policy themes are widely distributed, and the overall policy level is in a transitional stage from “average” to “good”, but there are imbalances in policy supply and quality among regions. Therefore, it puts forward specific suggestions for the early-developing regions and the catch-up regions. Additionally, measures to enhance the overall effectiveness of policies are presented. This is conducive to the high-quality development of the artificial intelligence industry.