Enhancing Erythemato-Squamous Disease Diagnosis: A Hybrid Approach with SVM SMOTE and Feature-Selection-Based Random Forest Model

The diagnosis of "erythemato-squamous" disease (ESD) is challenging, with complex symptoms and causes making it difficult to diagnose. Misdiagnosis can lead to serious harm to the patient’s health. This paper proposes a method based on SVM SMOTE and feature selection for the Random Forest model (RF). It utilizes SVM SMOTE to synthesize new data for the minority class of diseases, balancing the dataset, and selects important features to construct the RF diagnostic model. The model achieves a test accuracy of 99.50% and an average test accuracy of 98.62% in cross-validation, outperforming the SVM, NB, and DT models constructed in this paper. This paper provides a new, highly accurate approach for the diagnosis of ESD models, with the hope of addressing the model’s shortcomings in future research.