Logga in
Logga in

Advancing Railroad Traffic Prediction Using Neural Network Models

This thesis focuses on replacing conventional tree-based ML models with advanced neural network models for more accurate and robust traffic predictions.

Digitalization including using modern technology as Artificial intelligence to importance the transport system is a vital initiative.   

Thesis Description

One of Trafikverket’s core services is to provide reliable traffic information to all railroad travelers in Sweden. In the past years a pioneering AI project have curated the necessary data and evaluated several tree-based ML models at the task of producing better train traffic forecasts than what is available today to railroad travelers.

This master's thesis invites two students with a solid grasp of neural network breakthroughs like GNNs, Transformers, and LSTMs to embark on an exciting journey. Our goal is to transform railroad traffic prediction by embracing modern neural network models and to push the quality of railroad traffic information in Sweden, to the benefit of us all.

Work included

This thesis focuses on replacing conventional tree-based ML models with advanced neural network models for more accurate and robust traffic predictions. The dataset, spanning 2015 to 2022, has been preprocessed to create a user-friendly foundation for initial predictions. As we progress, there's room to integrate additional data.


Your task is to explore recent literature, select the most fitting neural network algorithm, and put it to the test. We want to see if these innovative models outperform traditional tree models in predicting railroad traffic.


Begin with a comprehensive review of the latest neural network advancements. Once you've selected an algorithm, adapt it to our railroad traffic dataset. Through careful experimentation and comparisons, you'll uncover the strengths of neural networks in this context.


Your workspace awaits at Trafikverket labs -innovation arena for transport system in Linköping, Gelbgjutaregatan 2.

Work Environment

We provide a comfortable environment with a powerful dual GPU setup (RTX 3090) on an Ubuntu system. You'll have access to essential tools like Python, Visual Studio Code, Jupyter Notebook, and TensorFlow.


If you're well-versed in Python and Linux, with a keen interest in Machine Learning and Neural Networks, you're the right fit. Familiarity with GNNs, Transformers, and LSTMs is a plus.

Your Partner

In essence, this master's thesis offers an exciting opportunity for two students to reshape how we predict railroad traffic. With advanced neural network models and a supportive environment, you'll contribute to modernizing transportation prediction while gaining invaluable experience.


Start 2023 when approved applicants to -2024 January Report presentation.


Your application shall contain information as follows.

  • A personal letter describing yourself, your education, competence and why you with thesis partner are the excellent candidates.
  • Your CV
  • Certificate showing which courses you have taken with the achieved results.  

Contact us as below and you will receive an email address to the supervisor to send your application.


Daniel Jakobsson


Telefon: +46 10-123 42 17

Mattias Tiger

Technical lead

Telefon: 013-28 69 20

Fredrik Lemon


Telefon: +46 10-123 65 40