Subject of this thesis is the development and the application of an instance based learning method for the prediction of traffic variables. The objective is to generate information that can be used for dynamic traffic management purposes. The method uses traffic variables based on real traffic data that are structured with respect to spatio-temporal attributes. To make a prediction, the current pattern (consisting of continually observed traffic variables of the past hour and calendar attributes) is compared to a historic database. The similarity of two patterns is calculated using a weighting function that considers spatio-temporal attributes of the respective features. The most similar patterns are used to calculate the prediction. The method is tested based on real data (local occupancy and traffic flow, travel time based on probe vehicle data and vehicle re-identification) and evaluated in comparison with other prediction methods.
«
Subject of this thesis is the development and the application of an instance based learning method for the prediction of traffic variables. The objective is to generate information that can be used for dynamic traffic management purposes. The method uses traffic variables based on real traffic data that are structured with respect to spatio-temporal attributes. To make a prediction, the current pattern (consisting of continually observed traffic variables of the past hour and calendar attributes...
»