A Hybrid Deep Learning Model for Skin Cancer Classification Using Elephant Herding Optimization

Abstract

Background:

Skin cancer remains one of the most prevalent malignancies worldwide, and its prevention and early detection play a crucial role in reducing morbidity and mortality. Recent advances in deep learning have significantly improved the accuracy of automated skin lesion classification. However, several challenges remain in designing a computationally efficient hybrid model that integrates multiple lightweight architectures.

Materials and Methods:

This work proposes a hybrid deep learning model that combines the MobileNetV3+ and ShuffleNetV2 architectures. To enhance model performance, the Elephant Herd Optimization (EHO) algorithm was employed to optimize key hyperparameters, including learning rate, batch size, and the number of epochs. Skin image datasets were collected from SkinDataSe and the ISIC archive, encompassing both benign and malignant cases. Preprocessing involved contrast enhancement and various data augmentation techniques such as rotation, scaling, flipping, and translation to improve model generalization. All images were resized to 224×224×3 pixels and normalized to [0,1] range. The dataset was then divided into 70% for training, 15% for validation, and 15% for testing.

Results:

The hybrid model demonstrated exceptional performance, achieving a training accuracy of 99.72% and an F1-score of 0.93.

Conclusion:

The findings of this study confirm the potential of the proposed hybrid deep learning framework, optimized via EHO, in delivering accurate, swift, and early diagnosis of skin cancer.