South Central Railways (SCR) has been leading AI and DA initiatives in the Indian Railways (IR) and has been nominated to set up a “Center of Excellence (CoE) for AI and Analytics” with the Indian School of Business (ISB) as a knowledge partner in this effort. As part of this initiative, this report surveys current practices and applications of AI and DA in the logistics and mobility domain.

Using AI-Based System for Identifying Identical/Similar Items Across Railways for Ease of Procurement

Indian Railways manages over 280,000 different items, stocked across 215 depots across the country. The items procured by the Indian Railways are broadly classified into stock (S) and non-stock (NS) items based on usage. Items used regularly are categorized as stock items with a planned procurement process coordinated by the central stores in each zone. On the other hand, non-stock items are purchased locally by field units (depots or divisions within each zone) to fulfill short-term and primarily one-off needs.  As a result, NS items have large variability in the wording of the item description across purchase instances and are not allotted common identification number. 

The non-standardization item description poses a challenge for identifying relevant purchases from the database available in Stores Management System. Non-standardization refers to the same or similar items described differently and assigned different identification numbers in the material database. As a result, human intelligence is crucial in interpreting the complete description text and identifying relevant keywords. To address this need and with a larger goal of unified inventory management across all zones of Indian Railways in the longer term, the current project aims to develop data analytics and Artificial Intelligence techniques to identify similar/comparable non-stock and stock items used across Indian Railways.

An Analytical Model to Recommend Special Trains Operations

The Indian Railways  is the largest rail network in Asia, with a total running track length of approximately 99,235 km covering 7,325 stations. In 2019-20, IR carried about 8,086 million passengers resulting in passenger traffic revenue of 50,669 crore1. The Indian Railways operates two kinds of passenger trains – timetabled (scheduled) trains and special (unscheduled) trains. Special trains are planned on an annual or seasonal basis to capture observed demand surges on specific routes, particularly during holiday seasons and festivals. In 2019-20, IR operated 663 special trains carrying 25 lakh passengers on 2831 trips, and South-Central Railways (SCR) alone operated around 500 of those trips. 

The current system of predicting when to run special trains has several limitations. In particular, the planners must anticipate the origin-destination pairs, routes, and dates to recommend special trains based on Almanac and past experiences. The current project aims to develop an analytical model that will help planners decide/recommend (a) the markets (origin-destination pairs and routes) to operate such special trains in, (b) when to run these special trains and timings (c) the expected revenue, and service efficiency, potential. (d) composition of the train (e) best pricing model among the existing options.