Press

Efficient algorithms for course planning in school systems

Scheduling classes at schools and other educational institutions often poses huge organisational challenges. Now, a new research project entitled “Course planning in modular school systems”, headed by Rubén Ruiz Torrubiano, Senior Lecturer at IMC Krems’ Institute of Digitalisation and Informatics, is looking to identify new ways to solve this problem by developing efficient algorithms. Funded by Lower Austrian research funding agency Gesellschaft für Forschungsförderung NÖ, the project is being implemented in cooperation with Untis GmbH.
 

At IMC Krems, project manager Dr Rubén Ruiz Torrubiano is researching efficient algorithms for course planning in modular school systems together with Dipl.-Ing. Andreas Krystallidis, BSc, research associate at the Institute of Digitalisation and Informatics.

Timetabling – an organisational challenge

School timetable planning needs to reflect a variety of different requirements and preferences. Certain lessons need to be scheduled on particular days and at particular times, classrooms with sufficient space are required, and clashes need to be avoided in class and teachers’ timetables as well as in room usage. Pupils’ preferences are also an important factor. The outcome needs to be a timetable that satisfies all practical requirements while also taking students’ individual course choices into consideration.

Research project aims to develop efficient algorithms

Under the three-year project at IMC Krems, project manager Rubén Ruiz Torrubiano and research assistant Andreas Krystallidis, BSc, both from the Institute of Digitalisation and Informatics, are carrying out research designed to develop efficient algorithms for timetable planning. “Due to the complexity inherent in this problem, working out optimal solutions just isn’t practicable. Instead, we are focusing on methods that can be used to quickly calculate effective solutions. Our aim is to develop innovative approaches by combining human timetabling expertise with mathematical optimisation methods and artificial intelligence,” explains Rubén Ruiz Torrubiano.

Artificial intelligence and machine learning to help unlock solutions

Artificial intelligence will be used in the project in order to blend the know-how of human timetable planners with powerful methods of mathematical optimisation. In particular, the research will concentrate on machine learning using combinatorial algorithms. This approach is not only promising in terms of finding an effective solution to timetabling problems, it is also a key topic in the context of artificial intelligence. With current machine learning methods, algorithms tend to learn blindly from raw data and draw correlations. However, this means that the algorithm does not learn properly about situations that fall outside the available data. In order to solve this problem, alternative methods are needed that help to strengthen and improve such algorithms to make them more effective and robust when it comes to learning from new situations.

Research methods and cooperation with schools

To help achieve this goal, experts from schools and companies who are involved in timetabling on a day-to-day basis will be consulted in order to determine the specifications for the algorithm. Then, the researchers will work on different models using an iterative approach, with a view to optimising the models by integrating increasingly sophisticated algorithms. The algorithms will be tested using publicly available data and compared with other methods in competitions.
The research findings are ultimately intended to feed into the development of improved products by Untis GmbH, and will also be made available to a wider audience. Untis’s products are used in many Austrian schools, which means they could have a major impact on the efficiency of timetabling in the country.

“With this project we want to make a significant contribution to improving timetabling and consequently help to make the work of timetable planners more effective through the combination of artificial intelligence and mathematical optimisation,” Torrubiano reports.