engleski

Text mining student reports

This paper is a preliminary attempt at gaining useful knowledge from student reports, but also at inspecting their predictive power on grades or course success rate. First a descriptive statistical analysis is done in order to get a basic understanding of patterns found in reports. Then a frequency word analysis and contrastive analysis are done in order to identify the most frequent words per class and those words that occur distinctively between different classes. Prior to the classification task, correlations between the length of the final report and the final grade, and between the scores obtained on each reported activity individually and the final grade are inspected. Finally, four different classifiers are evaluated on the task of classifying texts based on the type. The same four classifiers are used to determine the predictive power that reports have on course final grades.

contrastive analysis ; predictive power ; classification ; text mining ; word frequency

nije evidentirano