Smart transport planning in action: predicting passenger flow in real time

Context: rigid schedules in a rapidly changing city

Hyderabad, a dynamic metropolis of over 10 million people, is at the forefront of India’s urban mobility transformation. Its metro system, spanning 69 kilometers across three automated lines, serves more than 500,000 passengers daily. Despite modern infrastructure, transport planning in the city has relied mainly on fixed timetables established far in advance.

These timetables were built on thorough demand studies and long-term surveys, taking into account factors such as weekday versus weekend traffic, holidays and seasonal variations. However, in practice, transport affluence – the actual volume and distribution of passengers across the network – can vary sharply due to local events, weather or shifting work habits. This gap between planned demand and real-world conditions sometimes leads to two costly outcomes: trains running under capacity or becoming uncomfortably crowded.

Hyderabad’s experience reflects a broader challenge in cities worldwide: how to make intelligent transport systems responsive to daily fluctuations in demand.

Innovation: intelligent transport systems that learn and adapt

Enter Rider Predict: a cutting-edge AI solution developed by the Keolis team. This tool is specifically designed to enhance transport planning by providing accurate passenger flow prediction at a granular level. Rider Predict uses machine learning models trained on two years of metro ridership data, enriched with real-time information on weather conditions and local events. The result is a predictive engine that offers daily recommendations, showing where and when demand will spike, station by station every 15 minutes.

What makes this solution stand out is its operational focus. Instead of offering theoretical insights, Rider Predict integrates into a simple web application tailored for metro regulators on the ground. This enables real-time adjustments to schedules and train frequency, based on predictive insights into transport affluence rather than static assumptions.

In Hyderabad, the metro’s closed system also provided a particularly rich dataset, allowing Rider Predict to reach greater precision in its passenger flow forecasts.

Benefits: smarter urban mobility for operators and riders

The initial results of implementing Rider Predict in Hyderabad were both impressive and practical. During a three-month test carried out by the local team as part of Keolis’ Booster innovation program, the model reached 96% accuracy in forecasting crowd levels.

Take the example of Monday, February 3, 2025: based on Rider Predict’s recommendations, the system was able to operate with 22 fewer trains that day, without any drop in service quality. This translated to more than 21,000 empty seats that did not run, avoiding wasteful energy use, reducing staff workload and cutting operational costs.

For transit authorities, the tool offers a compelling way to reduce environmental impact while improving efficiency. Fewer unnecessary train trips lead to lower energy consumption and a smaller carbon footprint, which are key goals for sustainable urban mobility.

For passengers, the impact is equally significant. Better alignment between supply and demand means shorter wait times, fewer overcrowded trains and a smoother overall transit experience. By transforming rigid scheduling into a flexible, responsive system, Rider Predict helps elevate the comfort and reliability of urban mobility.

Beyond Hyderabad, the Rider Predict approach is being adapted to a range of use cases. At Mont-Saint-Michel, a major tourist destination in France, the system helps manage shuttle services, adjusting to fluctuations in demand driven by tides, weather and holiday traffic. In Bordeaux, Keolis is using a similar data-driven approach to predict how many passengers will be affected by service disruptions, enabling better prioritization of incidents and more effective passenger communication.

Next steps: from prediction to proactive decision-making

With Rider Predict already delivering strong results in Hyderabad, Keolis is now focused on expanding its capabilities even further. The roadmap includes integrating new layers of data, such as detailed exit information and behavioral patterns, to push prediction accuracy to the next level.

The team also plans to implement scenario-based tools, allowing operators to compare multiple options for resource allocation, balancing cost, environmental impact and rider comfort. Such tools will elevate Rider Predict from a forecasting instrument to a strategic planning assistant, making it easier for decision-makers to act quickly and confidently in a variety of operational contexts. Another area of focus is evaluating long-term passenger satisfaction. Keolis is working to measure how better forecasting and smarter planning improve the public’s perception of transit reliability and comfort over time.

Ultimately, Rider Predict represents the foundation of a new generation of intelligent transport systems: systems that not only observe and predict but can eventually suggest and even automate key operational decisions. The future of public transit lies in platforms that can adapt in real time to changing circumstances, collaborating with human teams to deliver safer, cleaner and more efficient mobility for all.

In cities like Hyderabad, where population growth and urban complexity are accelerating, Rider Predict is proving that with the right tools, transit networks can evolve from static systems into dynamic ecosystems that are responsive, data-driven and deeply human-centered.