The Indian Railways has successfully completed a large-scale trial of an Artificial Intelligence programme it developed to address a persistent, vexing issue: waiting lineups.
The AI-driven module ‘Ideal Train Profile’ was able to reduce the size of the waiting list by 5% to 6%, ushering in a new era for how the Railways distributes its enormous inventory of berths.
There were simply more confirmed tickets for travellers at the time of booking at the conclusion of the test.
The ‘Ideal Train Profile’ was developed by the Railways’ in-house software arm, Centre for Railway Information System (CRIS), and fed with information on approximately 200 long-distance trains, including the Rajdhanis.
The AI module analysed patterns such as how travellers booked tickets, whether origin-destination pairings were successful or unsuccessful, and the time of year. In addition, it analysed which seats were vacant for how much of the travel.
Engineers engaged in the trial stated that the combinations of “training data” that ‘Ideal Train Profile’ could analyse were “almost limitless” because the AI module contained data from the previous three years.
According to them, the module learned variable data sets or possible ticket combinations by partitioning a single voyage into the number of stops and analysing passenger behaviour throughout them.
“If a long-distance train has 60 stops, the AI has learnt around 1,800 possible ticket combinations. If there are ten stops, there are around 45 ticket options, and so on,” a senior Railway Board official who did not wish to be identified told Deccan Era.
According to those participating in testing, ‘Ideal Train Profile’ became online at the beginning of the Advanced Reservation Period, or 120 days prior to the departure of trains, in this case the end of January. The test included passenger reservation systems from seven different railway zones.
Before the May-June holiday season, when demand for confirmed tickets is at its peak and passengers are most dissatisfied with long waiting lists, the Railways was eager to test out the kinks “good enough,” according to the officials.
Rail Bhawan bureaucrats have generally conceded that a significant percentage of passengers abandon the Railways because the national transporter cannot give in advance what they seek: a confirmed ticket.
At the beginning of the AI experiment, Sunil Kumar Garg, Additional Member (Commercial) of the Railway Board, informed the General Managers of zonal railways of this in a letter.
Garg argued that the rising rivalry from aircraft in first-class long-distance travel and buses in short-distance travel was a matter for concern. Introducing more passenger trains to accommodate expansion in particular crowded areas has been difficult.
“The need for a sophisticated Passenger Profile Management-based seat-quota redistribution has been felt for quite some time in order to boost the occupancy and revenue of the existing reserved trains,” the letter states.
However, rewriting the previous norms of seat distribution, freeing up seat quotas, and determining the number of confirmed tickets based on origin-destination pairings has been impossible, especially in real-time.
Now that the ‘Ideal Train Profile’ study has been successfully completed, officials have expressed optimism.
Another Rail Bhawan officer, who wished to remain anonymous, stated that the Indian Railways uses 1 billion ticket combinations for all reserved trains.
“The training data this AI would absorb is incredible,” he added, adding that the module can produce an additional Rs 1 crore per train every year, according to informal estimates.
“Likewise, the more AI learns, the more accurate it becomes.”