Using data mining to explore the spatial and temporal dynamics of perceptions of metro services in China: The case of Shenzhen

Improving perceptions of service quality is a promising planning strategy for increasing the attractiveness and retaining the ridership of public transport systems. The widespread use of social media presents an opportunity to investigate the performance of transport services from the customer’s perspective. This study proposes a framework for integrating quantitative and qualitative analyses to investigate the perceptions of transport services by mining data from social media. We utilise Sina Weibo data related to the Shenzhen metro system to illustrate a text mining process categorised by semantic, spatial and temporal information. On the semantic front, in addition to identifying service attributes and sentiment polarity consistent with previous literature, we find attributes specific to the Chinese context. We also identify clear temporal variations among different service attributes by visualising the number of corresponding microblogs across varying time scales such as hours, days and weekdays. The spatial variations reveal five main clusters around central business districts and transport hubs, which produce the highest density of reports about crowdedness, waiting times, reliability and frequency. However, microblogs that report on perceptions of safety and personnel behaviour present a different spatial pattern. This research offers insights into the ways in which we can use social media data to identify key service areas for immediate improvement and to monitor and manage metro systems more effectively over the long term.