Predicting Dropouts Using Machine Learning Models

High dropout rates in high schools and universities have become a major complication in many countries following the upsurge of the Covid 19 pandemic. With the accuracy provided, this model should be held in high regard as it serves as an efficient method for predicting possible college dropouts in university.

High dropout rates in high schools and universities have become a major complication in many countries following the upsurge of the Covid 19 pandemic. A student’s experience at school is one of, if not, the largest contributors to the likelihood that they drop out. However, with the growing quantity of students in school, and the sheer variety of those who drop out, manually pinpointing possible dropouts becomes extremely challenging. Therefore, by using a machine learning model, administrators would be able to more efficiently identify a possible dropout and provide the necessary resources to ensure their success. The data, acquired from the UCI Machine Learning Depository, had 24 inputs, and 3630 samples. For this particular project, Logistic Regression, K-Nearest Neighbors, Naive Bayes, Extra Trees Classifier, and an MLP Neural Network models were tested. Among these models, the most accurate was the Extra Trees Classifier model, with a percentage accuracy of 77.41%, a loss of 7.85, and an F1 score of 82.35%. With the accuracy provided, this model should be held in high regard as it serves as an efficient method for predicting possible college dropouts in university.

Explore More!

Research Paper

Project Webpage