PERFORMANCE EVALUATION OF MULTILAYER PERCEPTRON ARTIFICIAL NEURAL NETWORK MODEL IN THE CLASSIFICATION OF HEART FAILURE

Abstract

Abstract Objective: The aim of this study was to compare the classification performance of heart failure using the MLP ANN model on an open-access “heart failure clinical records” data set, as well as to identify risk factors that may be linked to heart failure. Material and Methods: The open-access “heart failure” data collection MLP ANN model was used to classify nephritis of the renal pelvis, and risk factors that may be involved were discovered. Different output metrics are used to demonstrate MLP ANN's progress. Results: It has been shown that the classification of renal pelvic nephritis is quite high with MLP ANN model (AUC = 0.925, Accuracy = 93.9%, Balanced Accuracy = 89.2%, Sensitivity = 98.4%, Specificity = 80.0%). Furthermore, the MLP ANN model showed that “time” is the most significant variable among the risk factors linked to heart failure. Conclusion: Consequently, in the analysis with the heart failure data collection, the MLP ANN model generated very positive results. Moreover, this model has gained important information in identifying risk factors that may be associated with heart failure. Thus, it has been understood that the relevant model will provide reliable information about any disease to be used in preventive medicine practices.