HAM10000 Skin Disease Detection And Classification Using Deep Learning

This study uses the HAM10000 dataset to present an experimental method of classifying skin diseases using a hybrid model that combines CNN and K-Nearest Neighbours (KNN). The suggested model is trained and assessed using the HAM10000 open-source dataset that contains skin lesion images of different areas of the human body. The hybrid architecture makes use of the CNN algorithm that can extract and analyse the spatial features from medical images and KNN's ease of use. Careful data preparation, including appropriate dataset splitting and normalization, is required for the experimental procedure. Next, the hybrid model is trained using validation data, paying close attention to tuning parameters adjustments and performance tracking. Evaluation metrics, such as accuracy, precision, recall, and F1 score, gauge the effectiveness of classification models in assessing their performance. Confusion matrices and classification reports are used to analyse the response of testing the model on an independent dataset, offering insights into the model's advantages and problems. Our suggested KNN-CNN hybrid approach's efficacy is further confirmed by comparison with baseline models. Our proposed model makes a significant contribution to the field of dermatology by offering a thorough analysis of the classification of skin diseases and highlighting the significance of integrating deep learning architectures with conventional machine learning techniques for improved diagnostic accuracy. The results present important new directions for investigation and possible uses in automated dermatological diagnosis systems.