Retrospective Dropout Study in Colombia University Students - Dataset and Source Code

This repository contains the dataset, R source code, and supporting documentation for the study “Academic Trajectories as Predictors of Graduation and Dropout in Higher Education: Comparing Logistic Regression, Random Forest, and Artificial Neural Networks.” The study examines academic performance, progression, and sociodemographic characteristics of Industrial Engineering students at Universidad de la Costa, Colombia. The source dataset contains 3,994 student records, from which an analytical cohort of 2,734 students with definitive graduation or dropout outcomes was selected. Logistic regression, random forest, and an artificial neural network were compared using repeated stratified 10-fold cross-validation and a held-out test set. The R code covers data preparation, descriptive statistics, between-group comparisons, correlation and multicollinearity assessment, model development, performance evaluation, and variable importance analysis. An English README provides execution instructions, while an English data dictionary explains the original Spanish variable names, category coding, units, and derived variables. The analysis uses retrospective summaries of students’ observed academic trajectories. Its results describe the classification of graduation and dropout outcomes and should not be interpreted as validated early-warning predictions.