Live Train Delay Prediction Using Machine Learning Technique and an Investigation Using Amtrak Rail Schedules

Abstract

Train disruptions have a significant influence
on passengers' choice to utilize rail transit.
This research develops immediate form
commuter train disruption forecast Passenger
Train Delay Prediction (PTDP) predictions
using data mining methodologies. This
research investigates the impact on PTPD
systems using this oriented Real-time based
Data-frame Structure (RT-DFS) and this with
historically based Real-time with Historical
based Data-frame Structure (RWH-DFS).
The findings show that PTDP methods
combining MLP with RWH-DFS outperform
any other model. External factors including
history delaying profile at the point of arrival
(HDPD), traffic and humanity, day of the
week, topography, and climate information
are all investigated and studied in immediate
form PTPD systems. The ability of this
method to boost the precision for predicting
train arrivals, time delays is crucial for
airport development.
We must utilize a time serial database as an
output in our process. After that, we have to
include predictive methods like logistic
regression and random forest. The
experiments conducted show that the degree
of precision and error levels of each algorithm
varied. The system has excellent prediction
efficiency and can track the pattern of
numerous delay factors correctly.