Preventive crisis management through ML forecasts

Project description

This innovative project focuses on the development of preventive management systems to avoid crisis situations in public transportation. In contrast to reactive approaches, which are only activated after a problem has occurred, this system uses advanced machine learning algorithms and predictive analyses to identify potential disruptions at an early stage and initiate proactive measures. Particular focus is placed on the prediction of critical water levels on transport infrastructures as well as the prediction of delay patterns and their cascading effects in the network.this innovative project focuses on the development of preventive management systems to avoid crisis situations in public transport. In contrast to reactive approaches, which are only activated after a problem has occurred, this system uses advanced machine learning algorithms and predictive analyses to detect potential disruptions at an early stage and initiate proactive measures. A particular focus is on predicting critical water levels on transport infrastructure and forecasting delay patterns and their cascading effects in the network.

The system supports decision-makers in preventive planning in scenarios such as

• Rising water levels with potential impact on rail traffic
• Foreseeable weather-related infrastructure loads
• Early detection of developing delay patterns
• Identification of potential staff shortages before they occur
• Preventive reallocation of resources to avoid congestion situations

As part of the project, students are working with real operating data from Transdev to develop a transparent dashboard that visually displays decision aids and explains the underlying prediction models. This approach combines practical teaching with concrete benefits for the mobility sector and at the same time promotes the development of more explainable AI systems.

Project goals

• Development of ML models for precise prediction of critical water levels and their development over time
• Implementation of algorithms to detect and forecast complex delay patterns and their propagation effects
• Development of a transparent decision support dashboard with explainable AI components
• Integration and processing of multiple real-time data streams from weather, traffic and operational systems, Development of an early detection framework with automated recommendations for preventive measures
• Design of an evaluation system to continuously improve the quality of forecasts and the effectiveness of measures
• Evaluation of the economic and operational benefits of preventive versus reactive management approaches
• Creation of a user-friendly interface for seamless integration into existing operational processes

Project team

Students

• Vaibhavi Balbadri
• Catrina Carrigan
• Umer Farooq
• Luca Gyhr
• Kateryna Rusnyak
• Ganesh Sahu
• Joelle Schneemann
• Marius Schweitzer
• Jan Süßmann
• Jonas von Werne

Supervisor

• Mara Burger
• Jan vom Brocke