Explainable Machine Learning for Drug Classification

This article provides a machine learning-based drug categorization research effort. The public repository Kaggle is where the dataset for this study was obtained. Age, sex, blood pressure (BP), cholesterol, and the Na-to-potassium ratio are the feature sets with the medication type as the target feature. In this work, five machine learning methods were applied: CatBoost, LightGBM, extreme gradient boosting machine, and extra tree. The findings indicated that, except for extra tree, all four algorithms had 100% accuracy, with CatBoost doing the best. The training and testing performance of the models was displayed using the learning curve. The model performance and key characteristics were understood using explicable approaches like SHAP and feature permutation significance. The findings indicated that the most critical characteristics for medication categorization are age, sex, and blood pressure. This work sheds light on how to classify drugs using machine learning. The findings demonstrate that machine learning may be used to classify drugs with high accuracy. The study's usage of explicable approaches can aid in understanding the model's performance as well as the key elements that can be employed to enhance it.