Desain Sistem Inferensi Fuzzy Mamdani Dinamis Berbasis KKM Adaptif untuk Evaluasi Holistik Pembelajaran Matematika

Abstract

Integrating cognitive and psychomotor assessment under Indonesia’s Kurikulum Merdeka often produces interpretive uncertainty and subjectivity bias, particularly when teachers must convert scores of Mathematical Concept Understanding Ability (KPKM) and hands-on skills into a single representative final grade. Existing fuzzy-based systems generally rely on static membership functions that must be manually redesigned whenever the Minimum Mastery Criteria (KKM) policy changes. This study aims to design a dynamic Mamdani Fuzzy algorithm architecture based on adaptive KKM parameters to integrate KPKM and hands-on skill scores into a single Learning Success Level (TKB) score through an analytical simulation approach using four linguistic sets (Needs Guidance, Sufficient, Good, Very Good), sixteen implication rules, MIN-MAX inference, and Centroid defuzzification, tested through eight boundary cases and six representative input-combination scenarios. The results show that the membership functions remain mathematically consistent at every critical curve point, while the inference process yields crisp values consistent with the designed reasoning logic, including under imbalanced input conditions between the two ability domains. The model proves consistent and adaptive to changes in KKM without requiring formula modification, thereby contributing a foundation for a holistic evaluation instrument aligned with the authentic assessment principles of Kurikulum Merdeka.