Abstract:

Objective and qualitative assessment of learners‘ knowledge and skills is an important task. On the one hand, it is a measure of how the educational content is taught, and, on the other, it is an indicator of the learners' personal achievements. A fair assessment gives the learners accurate information about the level of their knowledge, skills and abilities, and increases the teachers’ satisfaction by proving their good judgment - they have neither underestimated nor overestimated the learner. In programming training, the final assessment is often complex and involves several components, that take part in its formation with different weights. Different forms and methods are used for testing and assessment - online or offline tests, open questions, writing, testing and reading code, problem solving, and more. In certain cases, in large courses, students are divided into groups and can be trained and assessed by different assistants who may have subjective influence on the final assessment. Determining and using dependencies between individual assessment components can reduce subjectivity in forming the final evaluation. The final assessment can be difficult to make if there are discrepancies between individual components that are mutually dependent. Besides, standard evaluation algorithms are not easy to create, and often the final result is determined on the basis of the teacher's own experience. The article presents an experimental study on assessment, based on neural networks, trained with the experience of a lecturer. The results of an experiment to assess the knowledge and skills in programming are presented.

Keywords:

Fair assessment, complex assessment, programming training, neural networks.