CLASSIFICATION OF SKIN LESION USING CONVOLUTIONAL NEURAL NETWORK: FINDING THE OPTIMAL EPOCH, HIDDEN LAYERS, AND HIDDEN UNITS

The prevalence of AI usage has exploded in the last decade. One aspect, image recognition software, has been heavily used in different fields of study, ranging from physics all the way to computational biology. In this study, the efficacy of image recognition software from TensorFlow and its parameters were tested and explored. After the specific parameters were divulged, a convolutional neural network (CNN) was used to classify images of skin lesions into seven different categories. The deep neural network implemented in TensorFlow achieved the average accuracy of 38%, which was significantly higher than a chance level of 14%. It was found that one hidden layer with 1000 hidden units at 15 epochs produced the highest performance compared to other simulation parameters.