An Assessment of Statistical Classification for Socially Oriented Learning Methodologies

Social networks based on mutual interest, affinity or leadership are spontaneously generated when the training activities are carried out through online learning systems wherein collaboration and interaction among participants is encouraged. The structure of those interactions, reflected in a network graph, is known to contain relevant statistical information about the dynamics of the learning process within the group, thus it should be possible to extract such knowledge and exploit it either for improving the quality of the learning outcomes or for driving the educational process toward the desired goals. In this work we focus on forums engagement, modeling forums' interactions as social graphs and studying the power of some of the graphs properties for success/ failure learning prediction. Our data source is a complete record of the activity of students in forums, collected over two consecutive academic years of a computer networks course at the undergraduate level. The results show that some of the measures under study are very good predictors of the students' performance.