Mukhammad Solikhin, Tiurma Lumban Gaol, Ike Fitriyaningsih, Frisda Sianipar, Richye Manik, Rosida Octavia Sitorus
In determining the type of skin disease, the role of a dermatologist is needed, this skin disease classification application is an alternative to classifying several types of skin diseases. This study uses box-counting to calculate the fractal dimension value of skin disease images and then classified using K-Nearest Neighbor (KNN). The first step of image pre-processing is removing the image background by thresholding then separating the channel into red channel, blue channel, and green channel. Furthermore, CLARE is applied to each channel to provide a clearer image. The last is to do canny edge detection to get an image with canny border lines. Then the image is processed using box-counting to get the fractal dimension features which are classified into 4 scenarios using red channel, green channel, blue channel and all channels with the training data scheme: testing, namely 90:l0, 80:20, 70:30, 60:40, 50:50. This study uses 20 K values in KNN to find the optimal K value. The highest accuracy is the 90:l0 scheme on the green channel of 72.5% with K=l5 as the optimum k value. Based on the results obtained, it is concluded that fractal analysis and KNN can be used for classification of skin diseases. © 2024 Author(s).
Universitas Negeri Malang, Malang, Indonesia; Lnstitut Teknologi Del, Toba, Indonesia
