A convolutional neural network for skin cancer classification

Nur Nafi'iyah, Anny Yuniarti

Abstract


Skin diseases can be seen clearly by oneself and others. Although this disease is visible on the skin, sometimes we worry if this skin disease is not mild. Some people experience skin diseases directly and quickly go to a dermatologist to have their complaints and symptoms checked. This skin protects the body, especially from the sun, so it can lead to death if something goes wrong. One example of a skin disease that can be deadly is skin cancer or skin tumors. In this research, we classified skin cancer into Benign and Malignant using the convolution neural network (CNN) algorithm. The purpose of this research is to develop the CNN architecture to help identify skin diseases. We used a dataset of 3,297 skin cancer images which are publicly available on the Kaggle website. We propose two CNN architectures that differ in the number of parameters. The first architecture has 6,427,745 parameters, and the second architecture has 2,797,665. With both architectures, the accuracy of the first model is 93%, and the second model is 74%. The first model with the number of parameters 6,427,745 We save for use in the creation of the website. We created a web-based application with the Django framework for skin disease identification.

Keywords


Architecture; Benign; Convolution neural network; Malignant; Skin cancer

Full Text:

PDF


DOI: http://doi.org/10.11591/ijict.v11i1.pp76-84

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

View IJICT Stats