The Spatiotemporal Patterns of Bus Passengers: Visualisation and Evaluation using Non-negative Tensor Decomposition

Spatiotemporal analysis of passenger mobility patterns provides valuable information regarding the travel behaviour of passengers at different spatial and temporal scales. However, in the spatiotemporal analysis of passenger mobility patterns, a few questions are yet to be answered: how does passenger travel behaviour change during different seasons? In developing countries like India where land use distribution is complex, do travel characteristics have a relationship with spatial regions of different land use? And what is the influence of people from nearby sub-urban and villages on the passenger mobility of urban areas if transit service is provided? Hence, this study developed a methodology to visualise and analyse spatiotemporal variations in the bus passenger travel behaviour among different spatial regions at hourly, daily, and monthly temporal resolutions using non-negative tensor decomposition (NTD). Six-month electronic ticketing machine (ETM) data of the Davangere city bus service is collected. Land use data is also collected from the urban development authority of Davangere city. NTD was found efficient in extracting spatiotemporal patterns. From the analysis, it is observed that passenger mobility patterns across different spatial regions varied during different seasons and within a season as well. Pertaining to spatial variations, passenger origins and destinations are aggregated with respect to spatial regions with uniform land use or similar travel characteristics without giving any geographical inputs. Also, the mobility pattern of sub-urban and village people varied unconventionally. Thus, developed research methodology has the potential of unveiling the spatiotemporal variations in passenger mobility, which can act as a base for designing transit facilities and framing policies.