A logistic regression-based model for predicting heart failure mortality

Recent trends in evaluating World Wide Web data include the use of traditional data mining techniques, such as regression, clustering, and classification. This paper aims to develop a model for predicting heart failure mortality based on a publicly available online dataset containing medical records of 299 patients. Since the prediction outcome can have only one of two possible values, the binary logistic regression technique was applied. Research shows that the predictive model created using logistic regression can accurately predict patient mortality based on their clinical characteristics and identify the most significant attributes among those included in their medical records. In addition, applying logistic regression ensures the simplicity and interoperability of the developed model, which was a major drawback of previous studies. The prediction model was created using the RapidMiner software tool. Its contribution lies in incorporating a broader range of clinical attributes, leading to a more comprehensive approach that enhances accuracy and prediction efficiency. The accuracy, precision, and sensitivity values of the developed predictive model are approximately 80%, confirming the model's high quality. The Area Under the Curve (AUC), which provides a graphical overview of the model's overall performance, is 86.7%, reflecting its effectiveness. The indicators of the developed model exhibit strong overall performance, creating the potential for its application to assist healthcare institutions in assessing the clinical status of patients with cardiovascular diseases.