Analisis Performa Algoritma K-Nearest Neighbor dalam Klasifikasi Tingkat Kerontokan Rambut

Abstract

Hair loss can lead to baldness and affect one's self-confidence. Normally, hair falls out at 80-120 strands per day, and the average number of hair follicles on the head is around 100,000. If the amount is reduced by 50%, it can be considered a disorder. Therefore, a classification of hair loss levels is necessary to determine appropriate actions. Previous study has shown that the K-Nearest Neighbor algorithm is capable of classifying various diseases. In this study, the Luke Hair Loss Dataset from the website kaggle.com, consisting of 400 data points, was used. To evaluate the method's feasibility, a confusion matrix was employed. The objective of this research is to analyze the performance of the K-Nearest Neighbor algorithm. Several scenarios were utilized, including testing the model before and after SMOTE oversampling, testing before and after data normalization, testing based on different K values, and testing with varying ratios of training and testing data. The results of this study indicate that the K-Nearest Neighbor algorithm achieved the highest accuracy value of 0.9853, precision of 0.9886, recall of 0.9833, and f1-score of 0.9856.

Keywords: Hair Loss, Classification, K-Nearest Neighbor, Performance Test, Confusion Matrix