IMPROVING THE QUALITY OF PRACTICAL TEACHING BY APPLYING ELEMENTS OF MACHINE LEARNING AND PREDICTIVE ANALYTICS

The standard practice in calculation and laboratory exercises often involves students working with the same dataset. While this simplifies the process, it also creates a significant opportunity for academic dishonesty, such as plagiarism, as students may resort to using others’ work or results. This undermines the intended learning outcomes. Additionally, there is tendency of the use of measurement data from real world which presents some challenges, particularly in terms of providing a large volume of diverse datasets due to limitations in time, funding, resources, and the accuracy and repeatability of measurements. This paper introduces a new approach that leverages machine learning and predictive analytics to address these issues. Starting with an initial real dataset obtained through experimental methods, different models are used to predict results, generating new synthetic datasets for educational purposes. By varying the models and their parameters, multiple versions of the data are created, ensuring that smaller student groups receive identical datasets for their tasks, while in some cases, each student may be assigned a unique dataset. The paper also has an evaluation of the results achieved through this approach in teaching both basic and master’s level courses in a one technical faculty’s study program.