Study on the Efficacy of a Novel Sedative Medication Based on Wilcoxon Rank-Sum Test and Multiple Machine Learning Models

Abstract

This study investigates the efficacy of a novel sedative medication compared to an existing drug using the Wilcoxon Rank-Sum test and multiple machine learning models. The Wilcoxon Rank-Sum test revealed statistically significant differences in petco200, petco2005, IPI005, and moaas005 indicators, primarily within the first 1 to 3 minutes post-induction. Basic information between the novel and existing drug groups was comparable, suggesting effective variable control and attributing differences to the novel sedative. Subsequently, various evaluation metrics including MSE, RMSE, MAE, MAPE, and R² were employed. Exploratory predictions using the Random Forest (RF) model yielded suboptimal results. After comparing the performance of RF, XGBoost, CatBoost, LightGBM, and SVR, a combination of the RF model and logistic regression was selected for regression predictions based on data type. Visualization of results demonstrated good predictive performance and mitigated overfitting.