How an algorithm helps to create lesson plans

Creating lesson plans for an entire year is a complex task. Artificial intelligence can support this and often even plan better than humans. This is shown by an optimisation algorithm developed by and Ateleris GmbH.

Organisations spend a lot of time and energy coordinating and planning internal and external resources. This includes creating shift and deployment plans for staff or dividing events and assignments among different rooms and locations. But also arranging procedures and processes so that they can be integrated as well as possible into the workflow. The human being should always be at the centre. If these resources can be used optimally, various added values are created for the staff, the clients and for the organisation itself. The more consideration can be given to extended aspects, for example the preferences and availabilities of the employees, the greater these added values can be.

AI plans better than humans

Together with, Ateleris GmbH has developed an AI optimisation algorithm and suitable input and data interfaces. This solution enables the planning team of to create and optimise the lesson and resource planning for the entire coming year with the support of an algorithm. First, the availability and preferences of the teaching staff, consisting of over 30 trainers, are entered and merged with the course offers, the course demand and the course rooms. This is then used to generate an annual plan that fits perfectly and meets as many requirements as possible. This approach complements the work of the planning team in such a way that, compared to completely manual planning, a better result can be achieved in less time. Achieving a high level of acceptance of the approach among the planning team is crucial. Therefore, the AI optimisation algorithm was embedded in the existing planning process in such a way that it complements and supports the work of the planning team as transparently as possible. The planning team continues to enter the course offers, the availability and preferences of the instructors as well as the information on the course rooms into the existing data entry system. The Excel file generated from this is read in by the AI optimisation algorithm, from which it creates variants of the annual planning. The most suitable one can now be checked and, if necessary, extended or regenerated.

Implicit knowledge is missing

The success of such an optimisation system does not (only) depend on the implementation of the algorithm and the suitable modelling of the optimisation problem, but is equally determined by the selected input masks and UX designs. The more diverse and complex the included (framework) conditions are, the more difficult it is to query them efficiently and in a user-friendly way. An example: Trainer Meier teaches the courses “Project Management”, “Organisation” and “Team Leadership”, is available Monday to Wednesday in Zurich and on Friday in Bern and would like to design 3 to 6 lessons per teaching day, with 10 to 15 participants each. She also prefers to give the lessons on “Organisation” and “Team Leadership” on different days. The latter demonstrates another difficulty: implicit knowledge of the planning team. This is, for example, knowledge about preferences of the teaching team, which is used in manual planning, but usually does not find a place on the input screen. Therefore, it is important to give the planning team the possibility to adjust or fix certain proposed plan elements and let the AI optimisation algorithm create an annual plan again. The goal must be to augment the planning team’s skills, not replace them.

AI planning saves costs

Technically, the AI optimisation algorithm is based on “mixed integer programming” (MIP) at its core. The basic idea is to model the problem to be optimised, in our case the annual planning, as a mathematical system of equations in which the various resources are linked with their framework conditions. The latter can be defined as hard (not to be violated) or soft constraints. The system of equations is then solved by a so-called “solver”, whereby an objective function is minimised or maximised. This could, for example, be as follows: “Minimise the total costs caused by room rent and wage costs”. During the iterative solution process, care is always taken that hard framework conditions are not violated, otherwise the solution is invalid. If soft framework conditions cannot be met, this one solution becomes “more expensive”, i.e. the objective function is considered less good. In summary, the AI optimisation solution for reduces the planning effort and thus the costs for the implementation of the individual courses, while at the same time meeting the preferences of the trainers and their clients.

Creative Commons Licence

AUTHOR: Laszlo Istvan Etesi

Laszlo Istvan Etesi is the Managing Director of Ateleris GmbH. Together with his team, he develops data science and optimisation solutions for public and private clients from a wide range of industries at home and abroad.

AUTHOR: Jonas Schwertfeger

Jonas Schwertfeger is Chairman of the Board of Directors of AG. He is responsible for the strategic development of the Swiss training provider and, in particular, drives forward digitalisation projects in vocational education and training.

Create PDF

Related Posts

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *