Prediction of Delays in Public Transportation using Neural Networks

Abstract

The project the authors of this paper are involved in is titled "System for intelligent realtime timetable optimization and monitoring". The objective is to develop a system being able to use delay-predictions for real-timedelay-monitoring, and in the long term, for a timetableoptimization in the range of train networks. The presented paper deals with the part of the system responsible for processing existing delays in the network to generate delaypredictions for depending trains in the near future. Therefore a rule-based system was developed, processing a set of predefined rules with the input of a specific delay in a deterministic manner, delivering a resulting delay-scenario as output. This rule-based system was used as a comparison to the specially developed neural network in order to evaluate the accuracy and faculty of abstraction of such an artificially intelligent component. An excerpt of the real train network of the Deutsche Bahn was the basis for this research, for simulation purposes we used the SNNS (Stuttgart Neural Network Simulator) [5]. At the end of this paper we can draw a conclusion in favour of the neural network, which is able to abstract from known delay constellations.

Train Dispatching Management With Data- Driven Approaches: A Comprehensive Review and Appraisal

“…ANN, as a basic ML method, learn from historical data to make predictions about future [113]. Peters et al [114] applied ANN to process existing delays abstracted from known operation data to generate delay predictions for depending trains shortly; this method performs well when predicting future (secondary) delays based on existing (primary) delays, and it outperforms the traditional rule-based method. Yaghini et al [115] also presented an ANN model with high accuracy to predict the delay of passenger trains in Iran; the comparison of the proposed ANN, decision trees, and multinomial logistic regression models confirm that the ANN model has high accuracy, low training time, and remarkable solution qualities.…”

Learning From Accidents: Machine Learning for Safety at Railway Stations

“…This method has achieved greater forecasting accuracy compared with artificial neural networks [25]. For timetable improvement and real-time delay monitoring in a range of real train networks of the Deutsche Bahn, a delay prediction system has been developed utilising a neural network [26]. For studying and analysing large volumes of data, ML methods are growing increasingly powerful for track condition prediction, therein achieving improvements in future railway safety and service quality [27]- [29].…”

Predictive models for influence of primary delays using high‐speed train operation records

“…Artificial neural networks (ANNs) have been widely used to predict delays in passenger trains. On the basis of data in the Deutsche Bahn, Germany (Peters, Emig, Jung, & Schmidt, 2005) and data from 2004 to 2009 in Iran (Yaghini, Khoshraftar, & Seyedabadi, 2013) an ANN was proposed to estimate passenger train delays. However, Marković, Milinković, Tikhonov, and Schonfeld (2015) indicated that SVR is more accurate for predicting passenger train arrival delays than ANN algorithms based on Serbian Railways data.…”