When Artificial Intelligence Meets the Mobility Database – Five Practical Ideas for Smarter Public Transport

Throughout this series, I’ve explained how Finland’s Mobility Database brings fragmented public transport information together into a single, standardised data foundation, and how combining it with Fintraffic’s wider data ecosystem creates entirely new opportunities for digital mobility services.

The next logical question is obvious:

What happens when artificial intelligence is introduced into that ecosystem?

Good AI starts with good data

Artificial intelligence does not eliminate poor data quality.

It amplifies it.

Machine learning models depend on data that is consistent, sufficiently comprehensive and preferably supported by historical observations. Considerable effort in AI projects is typically spent preparing data before model development can even begin.

The Mobility Database changes that equation.

First, it provides standardisation.

Public transport data is already normalised according to GTFS and related standards, significantly reducing the amount of preprocessing required before analysis.

Second, it creates historical continuity.

Every disruption, delay and operational change becomes part of a growing historical dataset that allows machine learning models to identify recurring patterns instead of analysing isolated events.

Third, it combines historical and real-time information.

Historical data teaches models what has happened before, while real-time information enables continuous analysis as conditions evolve.

Finally, the platform is designed to integrate with other transport datasets.

Road traffic, weather conditions, railway operations and maritime information can all enrich AI models, creating a much more complete understanding of the transport system than any individual dataset could provide.

Five practical opportunities

Artificial intelligence is not a single application.

Different use cases solve different problems.

The following examples illustrate where AI could provide practical value on top of the existing data infrastructure.

1. Predicting disruptions before they occur

Historical disruption data already reveals recurring patterns.

By combining those patterns with road weather, traffic conditions and seasonal variations, machine learning models could estimate where disruptions are likely to occur before passengers are affected.

Transport operators would gain additional time to prepare operational responses, while passengers could receive early warnings before beginning their journey.

2. A conversational journey assistant

Today’s journey planners typically require users to specify routes, destinations and departure times through search forms.

Large Language Models offer an entirely different interface.

Imagine asking:

“How can I travel from Oulu to Helsinki tomorrow morning, arrive before ten o’clock and minimise transfers?”

Instead of navigating menus, travellers could simply describe what they need.

Behind the scenes, an AI assistant would combine information from the Mobility Database and other Fintraffic services before presenting a personalised recommendation in natural language.

Although technically ambitious, the underlying components already exist.

3. Automatic anomaly detection

Transport systems continuously generate signals indicating that something is wrong.

A stop consistently reports incorrect predictions.

A route repeatedly deviates from its timetable.

An operator suddenly stops providing real-time updates.

Machine learning can identify these anomalies automatically, allowing operators to detect data quality issues long before they become visible to passengers.

This is not merely theoretical.

Fintraffic already applies machine learning in traffic counting systems to improve data quality, demonstrating that similar techniques can also support public transport datasets.

4. Forecasting demand

Passenger demand changes constantly.

Rush hours, holidays, concerts, sporting events and weather conditions all influence how people travel.

By combining historical transport data with external information such as weather forecasts, event calendars and school holidays, AI models could forecast demand well in advance.

Operators could allocate additional capacity where needed while avoiding unnecessary services during quieter periods.

Authorities would gain stronger analytical support for long-term service planning.

5. Automatic service performance reporting

Transport authorities spend significant effort producing reports about service quality.

How reliable were services?

Where did delays occur?

Which operators experienced recurring problems?

Generative AI could automate much of this work.

Instead of manually analysing datasets, authorities could receive regular reports written in natural language, highlighting trends, anomalies and areas requiring attention.

Among the examples presented here, this is perhaps the closest to practical deployment because it relies primarily on historical analysis rather than real-time decision-making.

What is realistic today?

Not every AI application is equally mature.

Automatic anomaly detection and service reporting are already well within reach. The required technologies exist today, and the necessary data infrastructure is now available.

Predictive disruption analysis and demand forecasting require larger historical datasets, additional external information and careful validation before operational deployment, but they are technically achievable.

Conversational journey assistants remain the most ambitious vision.

Although Large Language Models have developed rapidly, combining them with multiple real-time transport systems while maintaining reliability represents a substantial engineering challenge.

Across every use case, however, the underlying requirement remains the same.

Reliable AI depends on reliable data.

Infrastructure before intelligence

Artificial intelligence is often presented as the starting point for digital transformation.

Our experience suggests the opposite.

The real foundation is data.

Without standardised, validated and continuously maintained information, AI projects spend most of their time cleaning data rather than creating value.

The Mobility Database changes that starting point.

Instead of solving the same integration problems repeatedly, developers, researchers and transport authorities can build on a common, trusted data foundation.

That does not guarantee successful AI projects.

But it removes one of the largest barriers that has traditionally stood in their way.

Ultimately, the most important achievement of the Mobility Database may not be the platform itself.

It is that, for the first time, Finland has created the digital infrastructure required for the next generation of intelligent mobility services.