Enhancing Healthcare Decision Making through Predictive Analytics Using Hybrid AI Techniques

R Punithavel; T Ramesh; P.P Mithun; R Anitha

doi:10.34293/sijash.v13iS3-i1-Feb.10270

R Punithavel Assistant Professor, Department of Computer Applications, KIT-Kalaignarkarunanidhi Institute of Technology, Coimbatore
T Ramesh Associate Professor & Head Department of Information Technology, Bharathiar University, Coimbatore
P.P Mithun Department of Computer Applications, KIT-Kalaignarkarunanidhi Institute of Technology, Coimbatore
R Anitha Department of Computer Applications, KIT-Kalaignarkarunanidhi Institute of Technology, Coimbatore

DOI: https://doi.org/10.34293/sijash.v13iS3-i1-Feb.10270

Keywords: Support Vector Machine (SVM), K-NearestNeighbor(KNN), Logistic Regression, Random Forest, Performance Metrics, Accuracy, Precision, Recall, F-Measure, Confusion Matrix, Ontology Classifier, Diabetes Prediction, Machine Learning, Predictive Analytics, Big Data, Healthcare, Classification Algorithms

Abstract

High levels of glucose in the bloodstream lead to the development of diabetes, which results in frequent urination, increased thirst, and increased hunger. It is crucial to address diabetes promptly as untreated cases may lead to severe complications in various body organs such as the heart, kidneys, blood pressure, and eyes. Predictive analytics over big data is a challenging task, particularly in healthcare. However, it can aid healthcare practitioners in making quick decisions about patients’ health and treatment based on big data. The performance and accuracy of ML algorithms used in predictive Data analysis for predicting the occurrence of diabetes are compared and analyzed across various disciplines. In this study, different classification Computational methods, which may involve various algorithms, such as SVM, KNN, Logistic regression, and Randomforest, were considered, and their performance metrics such as Recall, FN Measure, Precision, and Accuracy were evaluated Derived from the confusion matrix. According to the experimental results, the SVM and ontology classifiers yielded the highest accuracy for diabetes prediction.