Prediction-based model for student dropouts using modified mutated firefly algorithm

Abstract

Academic database is considered as the heart and soul of every higher education institutions. This database contains a vast amount of useful information that is useful for analysis. Algorithms for machine learning play a significant role in mining academic databases and have been proven to be effective when applied in the academic field. Prediction models are made using relevant classification algorithms for dropout analysis. The success of the prediction model depends on the performance of the feature selection algorithm used for dimensionality reduction. The study utilized the Modified Mutated Firefly Algorithm (MMFA) as a dimensionality reduction strategy to enhance the accuracy of the prediction model for dropout analysis. The results of the simulation revealed that the Decision Tree (DT) classifier outperformed the Naïve Bayesian using the three UCI datasets. After the test of benchmark datasets, a students' cumulative dataset was used to come up with a predictive model for dropout analysis of Davao del Norte State College, Davao del Norte, Philippines. The results of the experiment confirmed that the MMFA+DT obtained an accuracy rate of 95.82%, while MMFA+NB only has 92.85% using 10-fold cross-validation.