An Intelligent Multi-algorithm Machine Learning System For Skin Disease Classification And Prediction

• Author(s):

  M.Gughan Raja | S.Afreen Reikhana | S.jeya Varshini

• Keywords:

  Skin Cancer Detection, Deep Learning, Dermatoscopic Image Classification, And Transfer Learning, Grad-CAM, Saliency Maps, Data Augmentation, Class Imbalance Handling, And HAM10000 Dataset

• Abstract:

  This Paper Proposes An Interpretable And Explainable Multi-class Skin Lesion Classification Model For The HAM10000 Dataset. Timely Detection Of Skin Cancer Is Important But Difficult Due To Class Imbalance And Uninterpretability Of Current Models. The Model Employs EfficientNetV2-L With Channel Attention For Improved Feature Extraction And Classification. An Effective Preprocessing Strategy With Data Augmentation And Smart Class Balancing Is Used. Progressive Training In Three Stages Enhances Generalization. Visual Explainable AI, Including Grad-CAM And Saliency Maps, Explains Predictions Visually. The Model Has 91.15% Accuracy And 99.33% AUC Across Seven Classes. The Solution Enhances Diagnostic Performance And Interpretability, Enabling Its Use As A Decision Support System.

---

Other Details

• Paper id:

  IJSARTV12I4105213

• Published in:

  Volume: 12 Issue: 4 April 2026

• Publication Date:

  2026-04-30

Download Article