Android Application for Skin Cancer Prediction Base on Machine Learning

Skin cancer is one of the most common forms of cancer worldwide, and early diagnosis is critical for improving treatment outcomes and patient survival. However, manual diagnosis through dermatoscopic examination is time-consuming and highly dependent on clinical expertise. This work presents a deep learning-based approach for automated seven-class skin lesion classification using the HAM10000 dataset. The proposed system employs a ResNet50 convolutional neural network backbone fine- tuned through transfer learning and optimized using Focal Loss to address severe class imbalance among lesion categories.

The dataset consists of 10,015 dermatoscopic images classified into seven diagnostic classes: actinic keratoses, basal cell carcinoma, benign keratosis-like lesions, dermatofibroma, melanoma, melanocytic nevi, and vascular lesions. Extensive preprocessing and data augmentation techniques, including resizing, normalization, flipping, rotation, affine transformation, and color jittering, were applied to improve generalization performance.

Experimental results demonstrate that the proposed model achieves competitive classification accuracy with meaningful recognition rates across both majority and minority classes. Con- fusion matrix analysis and per-class metrics show that Focal Loss improves sensitivity for underrepresented malignant lesions while maintaining strong overall performance. To enhance practical usability, the trained model was deployed as an interactive Flask- based web application capable of real-time prediction, confidence visualization, and risk-level reporting.

The developed system highlights the potential of deep learning for AI-assisted dermatology screening while emphasizing the importance of transparent uncertainty communication, privacy preservation, and clinical validation before real-world deployment.