In the internet age, passengers expect transport companies to predict departure times with a very high level of accuracy. Prediction based on historical data can help to considerably improve the prediction quality of trip announcements. Prediction based on historical data supplies top quality data. More accurate arrival time predictions are achieved by extending the prediction service in the control system.
Current data sources are taken into account in the prediction, in addition to vehicle’s current timetable adherence. This helps particularly with hindrances caused by congestion, accidents, route closures or weather-related delays.
The model is based on measured values in the AVLC which register all vehicle travel times. These times are then saved in a central online database and included in the prediction.
To obtain highly accurate predictions, all measured travel times are broken down into a 30-minute time grid to take account of the varying traffic situations in the course of a day. At the same time, long-term statistics are also generated, categorised according to days of the week, weekends and public holidays. The observed travel times are used to weight these values so that the predictions can be constantly adapted. There is also an additional compensation possibility for early times or delays. Prediction based on historical data is bringing about a considerable improvement in prediction quality. This solutions support clearly the attractiveness of public transport. In general terms, it can be said that prediction based on historical data generates a great added value for all transport companies and the passengers.
Also read our UITP Summit 2017 recap: http://www.trapezegroup.eu/news/uitp-summit-2017-recap