The Comparative Analysis of Machine Learning Models for Quality Assessment of Textual Academic Works

Grading students' written work such as papers, essay questions on exams, and laboratory reports can be a significantly time-consuming task, although becoming a common practice. One of the common problems that arise when evaluating text documents can be the low quality of the source data that does not comply with the formatting rules. This study uses a Machine Learning approach to evaluate the quality of student paper text documents. During the experiment, data were taken from several works for testing. The quality assessment indicators used in this study are the compliance of the explanatory note with the structural elements of the document. Several Machine Learning algorithms and methods are tested, such as k-nearest neighbors (kNN), Support Vector Regressor (SVR), and Random Forest, to find the best model suitable for estimation. The result of the comparison reveals that kNN shows the highest prediction accuracy while having advantages in the learning process. The learning results of all algorithms are clearly presented in tables and graphical illustrations.