Graph mining approach for modeling academic success

The prediction of the academic success of a student, the change of success according to causes and processes, and the examination of the consequences of this change are a general research topic that deals with many disciplines from different disciplines. Using the approach in this study, patterns of the subset of the data set were obtained by using methods of finding frequently repeated sub-graphs, which is one of the graph-based data mining methods in the modeling of academic success. It has been observed that these patterns increase the performance of the machine learning model in predicting the academic success of the students. Rather than limiting success to only educational features and abilities, it is aimed to bring together the other factors that affect success and look at them from a broad perspective. In this context, many data sets derived from the demographic and academic background of each student are expressed in sets of graphs, and the data set is enriched with the frequent repetition of these graphs. Firstly, the performances of the models formed by the classical data mining classification methods were examined and then the performance of the models created with frequent repeating sub-graphs enriched data sets were also examined and compared. The results show that classifications made with models created with frequently repeated sub-graphs of enriched data sets yielded better performance results.