Abstract:    Industrial policy plays a crucial role in leading the development direction and allocating resource elements of China′s NEV industry. However, the dimensions of existing studies on industrial policy change and characterization are relatively single. Therefore, this paper proposed a three-dimensional analytical framework based on the purposefulness, direction and orientation of industrial policies from the perspective of policy evolution. The 1783 NEV industrial policies enacted from 2007 to 2023 are analyzed based on the computational text quantitative analysis method. Unsupervised machine learning LDA topic modeling was used to portray the evolution process of China′s NEV industrial policies, and the supervised learning random forest algorithm was used to predict and classify the policy types and summarize the stage characteristics of policy distribution. The study found that: (1) China′s NEV industrial policy has gone through a process of change from the initial stage to the rapid promotion stage to the high-quality development stage, and the distribution of policies has transitioned from the initial fragmentation and imbalance to a more centralized and balanced state. (2) The policies are mainly selective, regional and market-oriented, but the policy changes show a trend of gradual transfer to functional, industrial and technology-oriented. (3) The process of policy change shows the characteristics of market access system tends to be mature, financial subsidies oriented to be precise, infrastructure construction is increasingly important, and the trend of informationization and intelligence is strengthened. This paper has expanded the theoretical perspective of policy change research, improved the analytical framework of policy tools, enriched the analytical methods of industrial policy change characteristics and laws, and it will provide some practical insights for industrial policy optimization.

Key words: new energy vehicle, industry policy, policy evolution, machine learning