Ensemble learning techniques to improve the accuracy of predictive model performance in the scholarship selection process

Abstract

Scholarship selection with big volumes of college student data in an university undoubtedly required a lot of resources and time. Besides the inefficient factor, there are also human-error occurred in the scholarship selection process. Error and risk can be reduced with ensemble learning approach. The different with another method is that usually research will only choose one algorithm or doing comparison to search the best algorithm. But in ensemble learning, some of algorithms called base learner combined to shape a new more-established model. With ensemble, more accurate result of the scholarship selection produced and also had the most minimum error value. In this research, there are two algorithm used as the base learner which are K-Nearest Neighbor (KNN) and Support Vector Machine (SVM). Experiment showed result of KNN with 91% of accuracy and SVM with at the rate of 75% accuracy. This base learner combined into an ensemble learning model using voting classifier. After last experiment, ensemble learning model succesfully created and produced the scholarship prediction result up to 100% of accuracy on training data. With streamlit, an application has been made which can automatically determine whether a student is accepted or rejected in the scholarship selection process. From the result, ensemble learning model can be used by academics as a Decision Support System (DSS) for determining scholarship recipients. This model can also be used as a development for institutions in the academic field.

Streamlining Business Operations through ERP Implementation

“…Ensemble Learning achieved 100% accuracy in prediction, demonstrating its effectiveness in outperforming individual models. Potential drawbacks include the need for extensive training data, complexity in integrating multiple algorithms, and deployment challenges for real-time predictions and scalability [9].…”