Data Science2023
Predicting a Skin Lesion's State
Machine learning classification project identifying skin lesions associated with skin cancer risk, using the PAD-UFES-20 clinical image dataset.
Methodology
Multi-Otsu thresholding and Snake Contour segmentation for preprocessing, followed by color/asymmetry/texture feature extraction and hyperparameter tuning via grid and random search.
Findings
The tuned Random Forest model achieved optimal accuracy, showing potential as a clinical diagnostic aid.
Tools & Methods
K-Nearest NeighborsRandom ForestLogistic RegressionAUC-ROC