Public Transport in Danish Rural Municipalities: Topic Modeling, Semantic Classification, and Citizen Share Prediction Using Crowdsourced Data

Crowdsourcing has become a key instrument for civic participation in sustainable mobility and transport safety planning, yet rural municipalities often struggle to ensure adequate public transport service levels. This study applies a crowdsourcing-based analytical framework to two Danish rural municipalities, Thisted and Vesthimmerland, to identify dominant mobility-related issues and estimate perceived public transport level of service (LOS) in 2025. Topic modeling was implemented using BERTopic to cluster 2024 and 2025 citizen reports, while a zero-shot semantic classification model was used to infer LOS categories. An adjusted probabilistic blending model integrated topic weights, semantic scores, and official 2023 statistics distributed across a five-level LOS scale. Results indicate that topics related to road infrastructure defects and delays are strongly associated with low service levels. In Thisted, the predicted share of “Low LOS” increases from 36.9% (official statistics) to 44.8%, while in Vesthimmerland it rises from 36.9% to 41.3%. Recurrent deficiencies such as defective markings and missing signs reflect spatial conditions potentially linked to perceived service underperformance. Despite dataset heterogeneity and classification limitations, findings demonstrate that citizen-generated data can reveal latent perceptions of transport accessibility, offering a replicable and low-cost framework to support targeted mobility interventions in small municipalities.