A system for forecasting student performance based on course evaluation

The current work presents software capable of simulating the results obtained by students enrolled in an undergraduate course through the results obtained previously in other subjects of the course. The students and subjects were modeled as nodes and links in a graph and the growing models used in complex networks simulations were applied. The created graph represents similarity indexes between the students. The growing models were capable of predicting the evolution of the graph and, consequently, the grades obtained by the students to some precision. The current work presents four graph models based on Barabási-Albert and small world models as a way to represent relationships between students and subjects. Each presented graph model identifies clusters of students and subjects in different clustering indexes, and in this current work is discussed the benefits, limitation and growing characteristics of each model. The results obtained brought a good precision in the simulated grades, and they also made a discussion about the students and the course characteristics possible.