A Text-Based Recommendation System Analysis Using a Hybrid Machine Learning Model

The KIP-Kuliah program is an Indonesian government program to increase access to higher education for prospective students from underprivileged families, in line with the 4th goal of the Sustainable Development Goals (SDGs) on Quality Education. However, the manual selection process causes inaccuracy in targeting KIP-Kuliah recipients and also low accuracy in the selection process so that a more objective and efficient data-based approach is needed. This study aims to propose a text analysis-based recommendation system with a hybrid machine learning approach to improve the accuracy, efficiency and objectivity of the selection of KIP-Kuliah recipients using Natural Language Processing (NLP) with Bag of Words and TF-IDF text representations to process applicant descriptions and five Gaussian Naïve Bayes algorithm models, Logistic Regression, Random Forest, Support Vector Machine and Gradient Boosting Machine. To overcome data imbalance and feature optimization, this study explores the use of SMOTE and PCA (Principal Component Analysis). The results showed that Logistic Regression with TF-IDF without SMOTE and PCA provided an accuracy performance of 80.82% followed by Random Forest with the same accuracy but higher execution time. These findings indicate that the combination of NLP and hybrid machine learning can improve the accuracy of KIP Kuliah recipient selection, thereby supporting a fairer, more efficient and quality education system and supporting the development of achieving SDGs targets.