Technological changes caused by the COVID-19 pandemic have led to a sudden digitization of education, where online classes and distance learning platforms have become part of everyday life. This caused the accumulation of a large amount of unused data in the databases of educational institutions and thus opened the question of the potential of mining the data for the benefit of these institutions. With the use of artificial neural networks, it is possible to discover patterns in data that provide insight into student behavior upon which educational institutions are able to make better decisions. Despite the fact that artificial neural networks are already widely used to solve various problems, they are very little present in education. Therefore, the goal of this work is to investigate the possibilities of artificial neural networks in education by conducting research on the prediction of students' final performance at the end of the third year based on their grades from the first year of study. The research was conducted on a total of three models of artificial neural networks, and in order to reach a better conclusion, a linear regression model was also used for comparison. Based on the square deviation of the values, the research showed that artificial neural networks can very well predict the final performance of the student, where it even surpassed linear regression according to the results.