| Abstrakt: |
In the modern era of the digital world, digital devices have taken over many aspects of our lives, and the healthcare industry is no exception. Clinicians are also increasingly migrating patient healthcare data from paper to electronic formats. With the advancement of recent technology, the automated Clinical Decision Support System (CDSS) aims to make better prognoses and improve the quality of healthcare delivery. The proposed system focuses on significant improvement to satisfy the expectation of patients and doctors. In this work, we propose a framework to classify different diseases and provide outcomes to the patient in the form of recommendations. We focused on preprocessing and feature selection techniques to improve the performance and quality of care. Firstly, data is collected and preprocessed using encoding categorical features, min–max scaling and removing null and duplicate entries. After that, a Rough Set Theory (RST) is applied to select highly relevant and nonredundant features to reduce the dimension of the datasets. Then, six popular machine learning classifiers viz., K Nearest Neighbors (KNN), Linear Support Vector Machine (LSVM), Radial Basis Function Support Vector Machine (RBF SVM), Decision Tree (DT), Random Forest (RF) and Naive Bayes (NB) were employed on four different medical datasets such as breast cancer, heart failure, post-operative patient data and thyroid collected from UCI repository. Performance was compared with proposed classifiers and existing state-of-the-art classifiers in terms of accuracy, precision, recall and f1-score. Our proposed classifier DT achieved better accuracy for breast cancer, post-operative patient and thyroid while KNN achieved better accuracy for heart failure. Hence, overall performance strongly suggests that the proposed framework may help medical experts to diagnose various diseases more accurately. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink