Özet:
In congested cities where the commuting time doubles during peak hours, it is crucial to identify every network problem. While it is costly to implement classic sensor technologies that are used for identification all around the network, using trajectory data of GPS (Global Positioning System) equipped bus fleet is suggested in literature. In this study, it is aimed to analyze the traffic behavior of the buses around the routes of these buses, and to express the effects of the selected parameters on the buses near bus stops numerically using the bus GPS data in Istanbul (IETT). The data consist of more than 5000 daily trajectory log files, including more than 25 million rows of location and time information during April 2016, of buses working in 12 bus routes which are selected for the most variety. The influence distance, wherein the buses affect traffic network while slowing down around the bus stops, are measured for each bus stop by using Fused Lasso method on the speed patterns of buses along the bus routes. Possible interruptions and the correlation of their distances to the bus stops with the influence distances are investigated for the surrounding network of 438 bus stops. M5P, random forest and extremely randomized trees models are created to predict the influence distances using the bus stop parameters. The models show that, although the passenger demand plays huge role on the influence distances of the bus stops, the other parameters such as change in the number of lanes and location of the traffic lights should be used to predict the influence distances. The influence distance for the bus stops varies from 36 to 174 meters, with an average value of 98 meters.
