Modeling Public Transportation Policy Using Macroscopic Social Media Data Mining

Transportation policies must be created by the government, especially in countries with high population expansion, transportation services are used more to meet daily necessities. Conventional surveys to gauge public opinion are costly and slow; social media offers a macro-level proxy that can complement official data. This study employs large-scale online data mining to build decision support for transportation policy. We collected 19,806 Indonesia-based Twitter posts referencing public transport, private transport, sustainable mobility, and electric vehicles. After preprocessing, we fine-tuned IndoRoBERTa for sentiment classification and applied Latent Dirichlet Allocation for topic modeling. The sentiment model achieved 81.17% accuracy, with precision, recall, and F1-scores all above 0.80. Positive discourse concentrated on private vehicles, public transit, multimodal travel, and environmentally responsible practices, with many users endorsing eco-friendly private cars. Negative discourse emphasized severe air pollution, frequently attributing risk to emissions from private automobiles in Jakarta. Translating these insights into policy, we propose expanding electric-vehicle charging infrastructure, implementing vehicle buy-back/retirement programs, establishing low-emission zones, and promoting biofuels. The results demonstrate that macroscopic social media analytics can surface actionable public preferences and pain points, enabling near-real-time monitoring to inform adaptive and equity-oriented transportation policies. This framework provides a scalable approach for governments in rapidly growing contexts to align service provision with community sentiment while advancing sustainability goals.