A Semi-Supervised Learning Approach for Predicting Student's Performance: First-Year Students Case Study

Students performance is an essential part of a higher learning institution because one of the criteria for a high-quality university is based on its excellent record of academic achievements. The first-year of the lecture is the student period of laying the foundation that will affect academic success because first-year plays an important role in shaping the attitudes and performance of students in the following years. In this study, a semi-supervised learning approach is used to classify the performance of first-year students in the Department of Mathematics, Universitas Indonesia. Student performance will be divided into two categories, namely medium and high. The sample in this study consist of 140 first-year students with 27 features. There are two processes used i.e. clustering and the classification process. In the clustering process, the data is divided into three clusters using K-Means Clustering and the Naive Bayes Classifier is chosen to classify it. The performance of the proposed algorithms is stated by accuracy, sensitivity, and specificity value i.e. 96%, 92.86%, and 100% respectively.