Despite skin cancer being the most prevalent form of cancer in the world, it remains one of the most challenging aspects to diagnose due to the various forms it may take. Commonly, diagnosis revolves around visual examination and histopathology, which have their own disadvantages related to subjectivity and high expertise. This study proposed an alternative approach by utilizing the effective and efficient deep learning model EfficientNetB3. In the current study, deep learning techniques are studied in the context of their use in skin cancer spot prediction and classification. Deep networks based on the advancements in CNNs, notably the EfficientNetB3 architecture, are applied and proved to enhance the generalization capabilities of machine-based systems in cancer recognition. For the EfficientNetB3 model to learn complex patterns, it is pre-trained on extensive datasets and then fine-tuned and applied with transfer learning techniques to the dermatological images' specific characteristics. To increase the competition between features and reduce overfitting, dense and dropout layers are incorporated. This experiment's proposed setup, which involved a Kaggle dataset containing images of malignant and benign skin moles, shows the high achievement of the system in predicting and classifying skin cancer. The performance statistics, specifically the training and validation losses and accuracies, indicated the system's reliability and strength. The quantitative results presented here suggest that the proposed model is superior to the previous ones and thus indicates the possible level of accuracy and efficacy in skin cancer diagnosis by the proposed method.
Keywords: Deep Learning Skin Cancer Detection and Classification, Convolutional Neural Networks, EfficientNetB3, Transfer Learning, Dermatological Image Analysis.
How to cite this article: Voggu S, Siddiqui S, Performance Analysis of Modified EfficientNetB3 for Automated Skin Cancer Classification. Int J Drug Deliv Technol. 2026;16(2s): 327-241; DOI: 10.25258/ijddt.16.327-241