Measurement: Sensors (Dec 2022)
HRV based feature selection for congestive heart failure and normal sinus rhythm for meticulous presaging of heart disease using machine learning
Abstract
Heart rate variability (HRV) for subjects of Congestive heart rate failure and normal sinus rhythm is structured to be subjected to turbulence studies. When checking the scaling property for NSR subjects, it indicates the presence of measured heartbeat control mechanisms, and concerning CHF subjects found the scaling property is partially lost. So the absence of scaling properties indicates heart rate variations. Various factors regularize the regular pulse of the heart. It is observing the methods for noticing the interaction between the thoughtful and parasympathetic sensory systems. The versatility of the heart to outside also inner boosts is reflected by the pulse changeability (HRV) diminished. HRV is a key indicator that shows if patients have negative HD results. Given the Time domain, frequency domain, and nonlinear and profoundly complex elements parameters for heart disease. It has restricted the measure's ability to precisely investigate the basic elements. In this examination, we propose a framework to detect HRV by feature selection methods that highlight relevant features, and complex elements and the best method. This paper uses HRV features and the dataset of HRV for 419 instances which is real dataset (Max Hospital mohali) to be taken and feature selection methods are to be used such as the filter method, Wrapper method, and embedded method to select the relevant feature set. This paper's implementation is done by using a Machine learning classifier, for example, Random forest (RF), KNN, support vector machine, Naïve Bayes, and Decision tree be utilized to assess the identification execution. The experiment's result is best for the filtering method in requisites of Accuracy, and F-measure. The resulting performance on the data available is better than previous feature selection models that were used. The current work is implemented on the Python Jupyter notebook, it enables the evaluation and provides excellent performance. The results for the current work the filtering methods are considered the best method for feature selection the average accuracy without without-FS is 86.3492064, the filtering method is 95.3968254, the Wrapper method is 94.9206348, and the embedded method 93.8095236., the average score of F measure for without FS is 83.7949952, Filtering Method 95.2867576, Wrapper Method 94.8161982, Embedded Method 93.4696154, the average score is AUC for without-FS 85.2,650,446, by Filtering method is 95.0550574, for Wrapper method 94.6734954, for Embedded Method 93.0729834, the basic approach of feature extraction is also shown for this current work. From the health point of view in this study, the HRV parameters using machine learning technologies could easily predict heart disease at early stages, and accuracy in results achieved higher.
