Skin Cancer Detection using LeNet Classifier in Comparison with Wide ResNet Algorithm with Better Accuracy

The goal is to increase classification accuracy by conducting a comparative analysis between the Wide ResNet algorithm and the LeNet classifier for skin cancer diagnosis. With the exception of melanomas and moles, each subset of the dataset gathered from the open access website Kaggle was split into an equal number of photos. The dataset was derived from the International Skin Imaging Collaboration (ISIC) dataset. Group 1's anticipated sample size is 819, Group 2's sample size is 819, and there are 1638 sample sizes overall.With power = (0.85) and alpha (α) = 0.05, the results for g power are 0.8. To gauge the accuracy, 70 percent images used as training data,15% are utilized for testing, 15% are used for validation and performance, for better performance. When compared to Wide Resnet methods, the research effort using LeNet for skin cancer diagnosis reached an accuracy of 95.56%. Using SPSS software, the independent sample t test was examined and found to have a significant result of 0.000, which is below the generally recognized significance level of p < 0.05. According to this perspective, LeNet algorithms seem to have a higher accuracy in the novel identification of skin cancer than Wide Resnet algorithms.