Implementasi Algoritma Fp-Growth Dalam Analisis Penerimaan Mahasiswa Baru

Abstract

Competition in higher education encourages universities to use historical admissions data as a basis for more focused recruitment decisions. This study implements the Frequent Pattern Growth (FP-Growth) algorithm to identify frequent itemsets and association rules in new student admissions data at Universitas Bumigora. The analysis used 4,858 applicant records with accepted status from the 2022-2024 admission periods and followed the CRISP-DM framework through the evaluation stage. Four combinations of minimum support and minimum confidence were tested: 5%/50%, 10%/60%, 15%/70%, and 20%/80%. The 5%/50% scenario produced the richest set of patterns, with 1,251 frequent itemsets and 4,763 association rules, while stricter thresholds substantially reduced the number of patterns. The data were strongly concentrated on the Mandiri admission pathway and accepted status. A more informative association linked applicants from vocational high schools (SMK) with the itemset {Mandiri pathway, accepted status, year 2023}, yielding confidence of 1.000 and lift of 1.624. These findings indicate that FP-Growth can support admissions evaluation by revealing segments that are difficult to see from descriptive summaries alone. The results can inform targeted outreach to vocational schools, evaluation of the Mandiri pathway, and enrichment of admissions attributes for future analysis.