Comparative Study to Find the Best Deep Learning Feature Extraction Models and Machine Learning Classification Algorithm to Detect Monkeypox Disease

Monkeypox disease that has spread throughout the world for the past few years has a form similar to other diseases, making it difficult to identify monkeypox correctly and patients end up getting the wrong or less effective treatment. Deep Learning is the right solution to detect monkeypox because of its ability to handle complex and varied data, such as medical images. With CNN architecture, Deep Learning can automatically learn deep features directly from data without manual extraction, making it effective in recognizing complex patterns that are difficult for humans to identify. This study aims to find the best Deep Learning and classification model to distinguish monkeypox from other similar diseases (acne, chickenpox, cowpox, hand foot and mouth (HFMD), and measles). The models tested are CNN models: VGG16Net, MobileNetV2, and DenseNet-201 which have been fine-tuned combined with the Random Forest classification model with input in the form of human skin images. Each model is evaluated by calculating the confusion matrix (accuracy, precision, recall, and F-1 Score) and AUC ROC to determine the best performance. The highest accuracy was produced by the MobileNetV2 and SVM models, reaching 94.41%, while the highest AUC ROC average result was produced by the DenseNet- 201 and SVM models at 98.67%. This study shows that Deep Learning models have great potential to help diagnose diseases accurately.