PENGELOMPOKAN KARAKTERISTIK MAHASISWA DENGAN ALGORITMA FUZZY C-MEANS DAN POSSIBILISTIC FUZZY C-MEANS

Abstract

Traditional learning with a one-size-fits-all approach often causes learning difficulties for students with various characteristics. This research aims to group students based on individual characteristics using the Fuzzy C-Means and Possibilistic Fuzzy C-Means algorithms, as well as identifying the factors that play a role in each group. This research uses data from Pertamina University students, including academic performance, educational background and living conditions. Principal Component Analysis was used for dimension reduction and identified 13 main components, including pre-college academic grades, parental education and income, as well as socio-economic factors. Clustering evaluation using Partition Coefficient, Partition Entropy, and Fuzzy Silhouette Index shows that Possibilistic Fuzzy C-Means has better performance in handling data with ambiguous membership levels. The grouping produces two clusters, namely outstanding students with the support of an educated family and independent students with a dominant learning style (Fuzzy C-Means), as well as outstanding students with the support of a highly educated family and independent students who are active in academic activities (Possibilistic Fuzzy C-Means).