Calculate chances of getting a confirmed ticket by analyzing booking and cancellation history

In India, every day millions of passengers travel by Indian railways. Passengers book their tickets and due to the sheer volume of passengers, the tickets get over very soon. Railways has a system of a issuing a Waitlisted ticket. If there are some cancellations, then the waitlisted tickets are confirmed.

The idea is that by analyzing the historical travel patterns, it will be possible to predict the probability of getting a confirmed ticket at the real time.

Here are some use cases:

Alok & Archana are planning to visit their parents on occassion of Diwali.
They try to book tickets but they are dissapointed becuase the tickets are already full and their booking status is WL50 (50 on wait list).
Alok uses the PNR Prediction utility to check what are the chances (probability) that WL50 will get confirmed. Alok finds that there is a 80% chance that WL50 will get confirmed by the date of his travel. Since there is a high probability that the ticket will get confirmed, Alok books the ticket.
Archana is really happy that they can celebrate Diwali with their parents.

Ashutosh and Poonam are planning to attend their friends wedding.
As ususal, the tickets are booked and the current booking status is WL60.
Ashutosh uses the PNR prediction utility and finds out there is only 40% chance that this WL60 ticket will get confirmed. Ashutosh sees that there is no point in booking this ticket and decides to travel by flight
Since Ashutosh is booking flight tickets in much advance, and gets a good discount and he also saves the cancellation charges and last minute dissapointments due to unconfirmed tickets in train.
Ashutosh and Poonam are excited to attend their friends wedding.

Not as simple as it sounds

It depends on lot of external factors. For e.g. are you travelling during a vacation time, or a festival time etc. We have developed an algorithm that can calculate the probability of getting a confirmed ticket.

Still experimental. Not accurate.

1. Obviously, this is still experimental.
2. The algorithm is not accurate. The algorithm can get accurate only over a long preriod of time. As and when the system knows about more and more data points, the algorithm will automatically become more and more accurate.
3. Right now, we support this algorithm only on limited sectors of travel. Gradually, we will support more sectors

By now, you might have realized that it is not easy to develop such algorithm / system. In order to run this experiment, we need bigger systems and more computing power.

If you feel that this experiment is worth continuing, and if you would like to support this project, then we request you to consider donating for this project.

You can support us by 2 ways:
• Donate directly to our paypal account