Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

SYNT: Neue Transportphänomene - Vom Maxwell-Dämon zum Verkehrsstau

SYNT 0: Neue Transportph
änomene – Vom Maxwell-D
ämon zum Verkehrsstau

SYNT 0.5: Vortrag

Dienstag, 27. März 2001, 16:00–16:15, S 7

Traffic Forecast Based on Real Data — •Roland Chrobok, Joachim Wahle, and Michael Schreckenberg — Physics of Transport and Traffic, Gerhard–Mercator–University, Lotharstr. 1, 47048 Duisburg, Germany

Different methods to forecast traffic are analysed from a physical point of view. For long term predictions heuristics are developed using a statistical analysis of historical data. Therefore, time series of daily traffic data from 350 inductive loops of the inner city of Duisburg over a period of three years serve as input. Daily and seasonal differences are discussed as well as the impact of special events, e.g., the solar eclipse in 1999. Additionally, commuter flows are identified.

For short term predictions two models are examined which take into account the current data: the No-dqconstantNo-dq and the No-dqlinearNo-dq model. The former uses a constant value for all horizons, e.g., single smoothing, whereas the latter extrapolates a linear trend into the future. A comparison yields that the constant model provides a better forecast than the linear. The results can be improved with a method that combines the short– and long–term predictions.