To develop a hybrid classification framework for the automatic detection of melanoma using Convolutional Neural Networks (CNN) and improved optimization techniques, specifically the Improved Gray Wolf Optimization (IGWO) method.
Key Findings:
The CNN/IGWO model achieved a test accuracy of 98.47% and an AUC (Area Under the Curve) of 98.2, indicating strong model performance.
The performance surpassed models using basic GWO and other state-of-the-art deep learning methods.
The IGWO mechanism improved convergence speed and classification performance.
Interpretation:
The integration of IGWO with CNN demonstrates significant potential for improving automated melanoma diagnosis, highlighting the importance of advanced optimization techniques in enhancing diagnostic accuracy.
Limitations:
The study may be limited by the dataset used (SIIM-ISIC 2020), which may affect the generalizability of the findings to other skin cancer types.
Potential overfitting due to high accuracy results needs further validation to ensure robustness.
Conclusion:
The proposed hybrid model shows promise for practical applications in skin cancer diagnosis, emphasizing the importance of advanced optimization techniques in deep learning.
