Cluster analysis of larger-scale discrete data with application to estimating dwell time of bus route

In this paper,we develop a model and an estimation approach for analyzing larger-scale discrete bus dwell time. the main interest is that we can enhance the reliability of measuring bus travel time through calculating dwell time of bus routes rather than single bus stop. to begin with,the conception of dwell time at a bus stop will be redefined under special conditions. secondly,the larger-scale discrete data of bus dwell time is collected from nanjing transit based on an automated vehicle location system. the descriptive statistic analysis shows that bus dwell time at different stops are stochastic and out of order. however,the fluctuation of dwell time in different bus routes happens regularly. to support this hypothesis,the large number of discrete data is classified into two stages with cluster analysis and outlier process respectively. based on the data process,two-stage hierarchical regression model is established to forecast dwell time of bus route. it is proved that estimation error of the calculation model is less than 10%. the comparison results indicate that the macro calculation model could meet the requirements of engineering application. © 2013 pakistan journal of statistics.