Detection of Malignant and Benign Skin Lesions using the Influence of Activation Function and Accuracy Analysis in Densely Connected Convolutional Network Compared over Convolutional Neural Network

Detection of Malignant and Benign Skin Lesions using the Influence of Activation Function and Accuracy Analysis in Densely Connected Convolutional Network Compared over Convolutional Neural Network

연구 분야: Artificial Intelligence

학회: 2023 Intelligent Computing and Control for Engineering and Business Systems (ICCEBS)

The aim of this study is to detect Malignant and Benign Skin lesions using Novel Densely Connected Convolutional and compare the accuracy with Convolutional Neural Network. The Novel Densely Connected Convolutional algorithm and the Convolutional Neural Network algorithm are two groups used in this study with a total sample size of N = 10 for prediction of skin cancer and its types like malignant and benign . The G power values are 0.8 with 80% confidence interval. The Novel Densely Connected Convolutional accuracy for predicting skin cancer for the activation function Sigmoid is 80.8%, ReLu is 77% and Softmax 53%. while the Convolutional Neural Networks accuracy for Sigmoid is 76.3%, ReLu is 48% and Softmax 76% . The statistical significance between the two groups, as shown by the T Test p<0.001(2-tailed) significance value. The Novel Densely Connected Convolutional method, which had higher accuracy than the Convolutional Neural Network algorithm, was found to be much more effective at detecting both malignant and benign skin cancer.