Prediction of University Students' Academic Achievement by Linear and Logistic Models

Abstract

University students' academic achievement measured by means of academic progress is modeled through linear and logistic regression, employing prior achievement and demographic factors as predictors. The main aim of the present paper is to compare results yielded by both statistical procedures, in order to identify the most suitable approach in terms of goodness of fit and predictive power. Grades awarded in basic scientific courses and demographic variables were entered into the models at the first step. Two hypotheses are proposed: (a) Grades in basic courses as well as demographic factors are directly related to academic progress, and (b) Logistic regression is more appropriate than linear regression due to its higher predictive power. Results partially confirm the first prediction, as grades are positively related to progress. However, not all demographic factors considered proved to be good predictors. With regard to the second hypothesis, logistic regression was shown to be a better approach than linear regression, yielding more stable estimates with regard to the presence of ill-fitting patterns. Keywords: logistic versus linear regression, prediction, credits, academic achievement, advance in careerSe estudia el efecto de dos tipos de factores sobre el rendimiento de estudiantes universitarios: variables académicas de rendimiento previo y variables demográficas, mediante modelos lineales y logísticos. El principal objetivo del trabajo es comparar los resultados obtenidos con ambas técnicas estadísticas, para determinar cuál de ellos es más adecuado en términos de ajuste y capacidad predictiva cuando se pretende explicar y predecir el rendimiento académico, en función de variables de rendimiento previo y factores sociodemográficos. Como medida del rendimiento a predecir se empleó el avance en la carrera. Las hipótesis planteadas son: 1) El avance está directamente relacionado con las calificaciones en materias básicas de primer año y con variables demográficas y 2) Los modelos logísticos son más adecuados que los modelos lineales, ya que presentan mayor capacidad predictiva. Los resultados permiten confirmar la primera hipótesis en su primera parte, ya que el rendimiento previo está directa y significativamente asociado al avance en la carrera. Pero se cumple de forma parcial por lo que se refiere al efecto factores demográficos. Con respecto a la segunda hipótesis, la regresión logística mostró ser más adecuada que la lineal, pues arroja estimaciones más estables en relación con la presencia de patrones de mal ajuste. Palabras clave: regresión logística versus regresión lineal, predicción, créditos, rendimiento académico, avance en la carrera

Predicting Student Academic Performance: A Comparison of Two Meta-Heuristic Algorithms Inspired by Cuckoo Birds for Training Neural Networks

“…Different approaches have been applied to predicting student academic performance, including traditional mathematical models [3,4] and modern data mining techniques [5][6][7]. In these approaches, a set of mathematical formulas was used to describe the quantitative relationships between outputs and inputs (i.e., predictor variables).…”

How Educational Data Mining Can Predict Students’ Academic Achievement

“…Those four predictors are Sex (male and female), type of high school (Medicine and Engineering), Type of middle school institution (Public and Private), and Location (Montevideo and Inland). The result for this study found that the students with better grades in the first year of faculty have less risk of curricular lag in the future (Ayan and Garcia, 2008). On the contrary, Bydzovska and Popelinsky (2014) investigated how Educational Data Mining can help predict weak and good students.…”

Factors Influencing Academic Performance of the Students at University Level Exam: A Literature Review

“…Cramer's V score [70] 21 Data mining [11], [80] 9 Logistic regression [35], [38], [61], [63], [64], [77], [79], [81], [82] 22…”