Automatic assessment of Java code

Abstract

In this study a tool was developed which marks automatically computer programs written by novices in Java. Existing open source static analysers were evaluated in the development. Metrics and marking schemes were developed to mark and grade the programming assignments automatically. A sample of 59 first-year programming students’ projects was selected for marking. The projects were marked manually and automatically. The results showed that there is a high correlation between manual and automatic marking for all criteria. The mean Kendall’s tau is over 0.75 indicating a very high level of correlation between manual and automatic marking. However, depending on the marking criteria, non-systemic variations were found.