How IRCTC’s new website got its new prediction service for waitlisted tickets

A screenshot of the prediction service on the IRCTC website

How the code was developed

The job of devising the algorithm, which involved writing the code from scratch, was given to the Centre for Railway Information Systems (CRIS), the software arm of the Railways, which also runs the ticket-booking website IRCTC.co.in.

The prediction service mines the historical data the Indian Railways has about its own passenger operations and booking patterns to predict the possibility of a ticket getting confirmed. An estimated 13 lakh tickets are booked every day against a reserved accommodation of 10.5 lakh berths across classes, and there was no clarity on the possibility of a ticket getting confirmed.

The algorithm devised takes into account 13 years’ worth of historical data to arrive at a “robust workable model”.

One of the major challenges the developers faced is that peak seasons and booking patterns of trains change every year, based on the Hindu festival calendar. For instance, Diwali, Dussehra, Durga Puja and Chhath holiday periods, which see a deluge of passenger traffic, take place on different days every year.

An official had told The Indian Express in 2017 that the service would say the probability of a ticket getting confirmed is purely indicative. The predictive service rolled out Monday does carry the disclaimer that the information displayed is based on current and past trends.

The disclaimer also says the probability shown may differ from the actual possibility of a ticket getting confirmed since it’s based on various factors.

This follows lengthy waiting lists, which keep decreasing as and when cancellations take place, but without any official way of knowing if the waitlisted tickets stand any chance of getting confirmed.