Data science and GIS-based system analysis of transit passenger complaints to improve operations and planning

Abstract

Transit user complaints support system resilience by serving as a data source for service improvements. This study shows how geographic information system (GIS)-based analysis, econometric models, and latent class analysis can improve system-wide understanding of passenger complaints. The analyzed dataset consists of 718 passenger complaints concerning the operation of municipal lines in Jerusalem as the study region. The analytical methods consists of GIS-based analysis and statistical modeling: mapping, recursive bivariate probit estimation, negative binomial model estimation, and latent class analysis. The GIS-based analysis showed that the spatial distribution of complaints changes over time as a function of service disruption type and geographical area. The recursive bivariate probit model results indicated that the most acute sources of frustration are service problem recurrence and monetary loss, with the former caused by overcrowding, delays and line cancellations. The negative binomial model results shows that the number of complaints increases with an increase in the passenger boarding to bus arrivals ratio. Latent class analysis reveals that, in terms of both prevalence and customer frustration, overcrowding delays and line cancellations are the most acute problems in the study region. The proposed interface between transit complaints and GIS databases can readily be implemented by transport operators and authorities.