Enhancing Dermatology Diagnostics: A CNN-RF Hybrid Approach for Skin Lesion Classification

It is aimed at presenting an advanced approach to using AI techniques to classify skin lesions. The method employs a CNN and RF classifier model since it combines deep and random learning approaches. The accuracy of the CNN-RF hybrid model was tested using a dermoscopic picture consisting of 3700 images of melanoma, benign naevus, and seborrheic keratosis. The model even confirmed it – the overall accuracy was established at 92%, indicating its effectiveness. 5, exceeding the accuracy of independent CNN and RF models, which got 89. 2% and 85. 6% accuracies, respectively. The hybrid model has a certain predictiveness due to its strong classification capacities, with KPIs such as 90% accuracy. This indicates a more than acceptable capture rate at 8% with a recall of 91. 6%, and F1 score of 91. 2%. By adding RF as the final stage of the classification system and removing any outputs it had made as a feature for the following stages in the chain, overfitting was alleviated, and performance on the validation set was improved. The structure of the proposed hybrid model was outstanding in terms of discriminative capability and the melanoma AUC of 0. In ROC (Receiver Operating Characteristic) analysis, 94 was obtained. This supports its possible application in clinically diagnosing cancerous skin lesions at a relatively early stage. The comparative research also reestablished the efficiency of the CNN-RF hybrid model over other approaches like the CNN-SVM at 90.1%. From the outcomes, the CNN-RF hybrid model can enhance the possibilities of correct diagnoses in dermatology for education.