Analysis of Patient Satisfaction through Interpretable Machine Learning Algorithms

This research study intends to reduce the features and predict whether the patients are satisfied with the service provided by the hospitals. The proposed system classifies top five features and give more accuracy using the machine learning algorithm. The existing system has a limitation that it requires an optimization solver and increases the computing work if the number of variables become large. The proposed system considers 17 attributes in the dataset and five features are selected to evaluate the system to increase the efficiency. Since the correlation of several dataset features is nearly equal, they are eliminated. Chi-square test is one of the most efficient feature selection method to reduce the unwanted data or unwanted features from the dataset before training and testing the model for attaining better accuracy and reducing the complexity of the model. The taken dataset is imbalanced, it affects the accuracy, so SMOTE technique is used to balance the dataset. The acquired dataset is cleared of any potential irregular data and pre-processed with several methods followed by feature selection and model building. The SVM, Random Forest, XGBOOST and Ensembling of Random Forest and XGBoost are the classifiers that were employed. When using a machine learning approach for both training and testing, Random Forest ultimately has higher accuracy compared to other algorithms. This method has the amazing capacity to increase categorization and forecasting precision.