Research on Development Decision-Making for Shared Autonomous Vehicles Based on Social Media Data Mining

The successful deployment of shared autonomous vehicle (SAV) technologies is of great significance for improving road traffic safety and road capacity. Nevertheless, the practical implementation of SAV technologies is hindered by ambiguous user acceptance intentions and complicated multi-stakeholder decision-making processes. Existing studies mostly conduct static analyses of user willingness with fragmented variable selection, lacking integrated frameworks that integrate public opinion mining, behavioral modeling, and evolutionary game theory. Different from existing studies that predominantly rely on static questionnaire-based willingness analysis with subjectively selected variables, this study integrates social-media text mining, an extended Theory of Planned Behavior, and a tripartite evolutionary game model to jointly investigate user psychological mechanisms and multi-stakeholder dynamic interactions. To explore the long-term development trajectory of SAVs, this study crawls 23,943 short-video comment datasets. The Latent Dirichlet Allocation (LDA) topic model is adopted to preliminarily identify eight core factors affecting users’ adoption of SAVs, based on which a questionnaire survey on SAV adoption intention is designed. A total of 724 valid questionnaires were collected. Expanding the classic Theory of Planned Behavior (TPB), this study incorporates additional latent variables including perceived risk, employment impacts, and government attitudes to construct an extended TPB model, which quantifies the correlations between latent constructs and adoption intention. The empirical results reveal that government attitudes (β = 0.497, p < 0.001) and employment conditions (β = 0.434, p < 0.001) exert significantly positive effects on perceived risk. Perceived ease of use indirectly determines adoption intention through three mediating paths: attitude toward behavior, subjective norm, and perceived behavioral control. Furthermore, stakeholders are categorized into three groups (users, enterprises, and governments), and a tripartite evolutionary game model for SAV stakeholders is established to derive evolutionary stable strategies (ESS) via numerical simulation. The simulation results indicate that the strategy combination of economic incentives and small-scale pilot programs (adopted by governments) paired with the adoption strategy (adopted by potential SAV users) constitutes the evolutionary stable strategy. Variations in governments’ economic incentive costs and small-scale pilot operation costs significantly alter the decision-making behaviors of potential SAV users. This study integrates behavioral modeling and game-theoretic analysis. It delivers theoretical foundations for research on SAV adoption behaviors and provides quantitative decision-making evidence for governments to formulate phased incentive policies and for service platforms to control pilot operation costs.