9th International Conference on ICT for Sustainable Development

Authors - Hemal Patel, Premal Patel
Abstract - This research study uses a dataset of 1800 benign and 1497 malignant mole photos from the ISIC Archive to demonstrate a Convolutional Neural Network (CNN) approach for the early identification of skin cancer. Acknowledging the vital importance of visual diagnosis in the identification of skin cancer, the study endeavors to enhance automated classification through the utilisation of a deep learning model. Important phases in the 14-step process include loading data, categorical labeling, normalization, and creating models with Keras and TensorFlow backend access. Because the dataset is balanced, accuracy evaluation is possible, and the result is a commendable 97% accuracy and precision score. The research highlights the potential practical utility of the created model, while downplaying the significance of early diagnosis in skin cancer. The incorporation of Ensemble Technique architecture is also investigated, which improves the performance of the model even more. This thorough method shows how CNNs can effectively classify skin lesions visually and high-lights the potential of automated systems to support prompt and accurate skin cancer diagnosis.