Review of techniques, tools, algorithms and attributes for data mining used in student desertion

Abstract

This article makes a review of the application of data mining in the academic desertion of the students; with the aim of finding common elements used by different authors about desertion. The search of the articles was carried out in digital libraries, indexed journals, institutional repositories among others. The selection criteria were based on the depth of the techniques, algorithms, tools and attributes used in the publication. Among the results we have that most of the researches are related are supervised learning, we have the classification as the most applied technique and we use the algorithm j48 and the variable or class to predict as the academic state.

Comparison of Predictive Models with Balanced Classes Using the SMOTE Method for the Forecast of Student Dropout in Higher Education

“…After going through a series of phases and tasks from the CRISP-DM methodology, eight classification algorithms corresponding to three techniques (decision trees, decision rules, and Bayesian networks) were analyzed. In this research, the classification technique was considered because it is the most widely used in data mining, coinciding with the results obtained by [30]. It was shown that Random Forest, which corresponds to the decision trees technique, is the best classification algorithm for the prediction of university student dropout, coinciding with the findings of [25,31].…”

“…However, other studies like [8,33,43] ranked Random Forest as the best ranking algorithm for predicting student attrition in the university ecosystem. Furthermore, [31] highlighted the use of WEKA as the most widely used data mining software tool in the studies collected, coinciding with the research of [30]. This tool was also used for the predictive purposes of dropout by [12,21,33].…”

“…In [30], a review of the techniques, tools, algorithms, and attributes for data mining used in student dropout was developed. This research filtered those articles with common elements to detect dropout in the academic context.…”

“…The selected attributes are AgeIncome_U and sex. The consideration of the social factor in comparative studies and in the evaluation of predictive models coincides with [8,12,13,30,33,43].…”