Safitri, Rizki Alifia and Hidayati, Rahmatina (2024) Komparasi Metode K-Nearest Neighbor dan Naïve Bayes untuk Mengklasifikasi Resiko Diabetes. SMATIKA : STIKI Informatika Jurnal, 14 (2). pp. 297-303. ISSN p- ISSN: 2087-0256; e-ISSN: 2580-6939

Preview	Text Komparasi Metode K-Nearest Neighbor dan Naïve Bayes untuk Mengklasifikasi Resiko Diabetes.pdf Download (323kB) | Preview
Preview	Text Hasil Cek Plagiasi_Komparasi Metode K-Nearest Neighbor dan Naïve Bayes.pdf Download (1MB) | Preview

Abstract

Diabetes mellitus is one of the fastest-growing health problems in the 21st century. One of the causes is the lack of public awareness for regular health check-ups, while the lifestyle being led is quite unhealthy. Hemoglobin A1c (HbA1c) examination is highly recommended to detect diabetes. However, this service is not yet available at Posbindu in Bulupitu Village. Therefore, another approach is needed to detect the risk of diabetes early, namely through data mining. The data mining methods used in this research are the Naïve Bayes and kNN classification methods. The variables to determine the risk of diabetes include gender, age, family history of diabetes, frequent urination, Body Mass Index (BMI), blood sugar levels, and diabetes risk output. The division of testing and training datasets uses cross-validation and ratio (60:40, 70:30, 80:20, and 90:10). The best accuracy of the Naïve Bayes method was obtained by dividing the dataset using k-fold crossvalidation with k=2, achieving 96.1%. In the kNN method, the best results were obtained from the 80:20 dataset ratio. Manhattan distance was found to be the best distance calculation in this study compared to Euclidean distance and Chebyshev distance

Actions (login required)

View Item