Open AccessEducational DataMining: A Study of the Factors That Cause School Dropout in Higher Education Institutions in Brazil
Author(s) -
Marília Nayara Clemente de Almeida Lima,
Roberta Fagundes
Publication year - 2020
Publication title -
renote
Language(s) - English
Resource type - Journals
ISSN - 1679-1916
DOI - 10.22456/1679-1916.105950
Subject(s) - lasso (programming language) , context (archaeology) , dropout (neural networks) , linear regression , evasion (ethics) , investment (military) , stepwise regression , regression , econometrics , computer science , actuarial science , statistics , economics , mathematics , geography , political science , machine learning , medicine , immune system , archaeology , politics , world wide web , law , immunology
Context:In Brazil, there is a high dropout rate in higher education institutions. Thus, it is clear that evasion is a frequent problem and that it is necessary to analyze the factors that cause it to enable solutions that can mitigate/ reduce this problem. Objetive: (1)perform a correlation analysis (Pearson and Spearman) of the educational factores of the School Census; (2)propose school dropout prediction models taking into account educational and economic factors using regression methods (linear, robust, ridge, lasso, clusterwise regression). Methodology: used the phases of the CRISP-DM methodology. Results: the factors related to not allowing financial assistance are related to as evasion, namely: food, permanence, didactic material, transportation. There are also factors related to the study period. The regression robust and linear regression show fewer errors. Conclusion: the correlations used present the selection of factors in a similar way, thus following a linear distribution. This study can help to create more investment in public policies, as it ratifies factors are related to this dropout problem.