Comparison Of Bisecting K-Means And K-Medoids Algorithm In Grouping Junior High School Students Based On Bebras Challenge 2023 Results

Improving the quality of education, especially Computational Thinking (CT) skills, is an important priority in the digital era. Bebras Challenge, an international competition to measure Computational Thinking. The results show the diverse abilities of Junior High School students, so that appropriate mentoring methods are needed. This study compares the Bisecting K-Means and K-Medoids algorithms to group students based on the results of the 2023 Bebras Challenge, along with their Indonesian, Mathematics, Science scores, and the duration of competition preparation. The study was conducted through data preprocessing, application of clustering algorithms, and model evaluation using Silhouette Score at the number of clusters k = 2 to k = 10, and running time. The results show two main groups, namely the first group of students with high understanding, fast processing time, and effective preparation. While the second group of students with low understanding due to less than optimal preparation. Bisecting K-Means showed the best performance with a Silhouette Score of 0.623 and an execution time of 0.005 seconds at k = 2. This study provides insights for educators and policy makers to design more effective data-based learning strategies.