A simulation based forecast algorithm for public transportation

Abstract

This thesis describes a simulation based system for forecasting arrival and departure times in complex public transit systems. The variables included in the calculations are: time of day, vehicle schedule adherence, estimated passenger count on stop, interactions with other vehicles and vehicle position from both historical and contemporary data. The use of a simulation based algorithm simpli es the implementation of a complex model with a great number of di erent dependencies. Additionally, creating a forecast using historical data while looking at di erent data features to nd similarities gives a more accurate forecast than just looking at a moving average using data from the current day.