The Evolution and Prediction of Bus Transportation in Greater Sydney Using Econometric Models

Abstract

This thesis utilises econometric methods in the context of bus network service prediction utilising the Greater Sydney bus network between 1926 and 2013. Using historical bus GTFS data, the method with which this is transformed to find the level of service per link, as given by the Open Street Maps network is also shown. Weighted spatial variables are described where the strength of the spatial relationship is given by a region-level correlation matrix, also described within this work. The lagged service variable is found to define to a high degree the number of services experienced on a link in any given year, with the addition of complementary and competing spatial variables improving the model fit marginally or leaving it unchanged. As expected, lagged complementary variables have positive correlations with to service levels in the proceed- ing year, while competing links show the opposite relationship. The lagged service level model for the entire Greater Sydney region is further compared against the region-level spatial model, showing that only few circumstances offer a superior performance of regional models to the aggregated.