Analysis of machine learning strategies for prediction of passing undergraduate admission test

Abstract

This article primarily focuses on understanding the reasons behind the failure of undergraduate admission seekers using different machine learning (ML) strategies. An operative dataset has been equipped using the least significant attributes to avoid the complexity of the model. The procedure halted after obtaining 343 observations with ten different attributes. The predictions are achieved using six immensely used ML techniques. Stratified K-fold cross-validation is mentioned to measure the expertise of proposed models to unsighted data, and Precision, Recall, F-Measure, and AUC Score matrices are determined to assess the efficiency of each model. A comprehensive investigation of this article indicates that the resampling strategy derived from the combination of edited nearest neighbor (ENN) and borderline SVM-based SMOTE and SVM model achieved prominent performance. Additionally, the borderline SVM-based SMOTE and the Adaboost model performs as the second-highest performing model.

Keywords

• Machine Learning
• Balanced Dataset
• Adaboost
• Support Vector Machines (SVM)
• Precision