Data prediction for cases of incorrect data in multi-node electrocardiogram monitoring

International Journal of Electrical and Computer Engineering (IJECE) Vol. 12, No. 2, April 2022, pp. 1540~1547 ISSN: 2088-8708, DOI: 10.11591/ijece.v12i2.pp1540-1547  1540 Data prediction for cases of incorrect data in multi-node electrocardiogram monitoring Sugondo Hadiyoso1,2, Heru Nugroho1,2, Tati Latifah Erawati Rajab1, Kridanto Surendro1 1 School of Electrical and Information Engineering, Bandung Institute of Technology, Bandung, Indonesia 2 School of Applied Science, Telkom University, Bandung, Indonesia Article Info ABSTRACT Article history: The development of a mesh topology in multi-node electrocardiogram (ECG) monitoring based on the ZigBee protocol still has limitations. When more than one active ECG node sends a data stream, there will be incorrect data or damage due to a failure of synchronization. The incorrect data will affect signal interpretation. Therefore, a mechanism is needed to correct or predict the damaged data. In this study, the method of expectationmaximization (EM) and regression imputation (RI) was proposed to overcome these problems. Real data from previous studies are the main modalities used in this study. The ECG signal data that has been predicted is then compared with the actual ECG data stored in the main controller memory. Root mean square error (RMSE) is calculated to measure system performance. The simulation was performed on 13 ECG waves, each of them has 1000 samples. The simulation results show that the EM method has a lower predictive error value than the RI method. The average RMSE for the EM and RI methods is 4.77 and 6.63, respectively. The proposed method is expected to be used in the case of multi-node ECG monitoring, especially in the ZigBee application to minimize errors. Received Apr 29, 2021 Revised Sep 11, 2021 Accepted Oct 10, 2021 Keywords: Expectation-maximization Incorrect data Predict Regression imputation Root mean square error This is an open access article under the CC BY-SA license. Corresponding Author: Sugondo Hadiyoso School of Electrical and Information Engineering, Bandung Institute of Technology School of Applied Science, Telkom University Bandung, Indonesia Email: 1. INTRODUCTION Nowadays, there are advanced progress in applying computing technologies [1]–[8], that have significant progress in artificial intelligence. The development of wireless communication media on the internet of things application is always followed by the development of protocols to support multiple or multiuser access. Multiuser monitoring or control applications have been applied in one of them in the health area. This application allows for centralized, fast, easy, remote, and multiuser health monitoring. Health parameters that get serious attention are the heart of this refers to the risks posed if not maintained optimally. Observation of heart conditions can be done by studying the electrical activity of the heart through an electrocardiogram (ECG) [9]–[11]. Previous research by Hadiyoso and Aulia [12], has succeeded in designing and implementing an ECG monitoring system for several ZigBee-based user nodes. But in its application, there are crucial problems, namely damage or loss of data if more than one active node is sending data streams [13]. This problem is likely to occur because of the failure of synchronization between the user/end node and the coordinator. Estimating missing data is a significant advancement that occurs during the data cleaning stage. Numerous studies have demonstrated that improper data management results in inaccurate analysis [14]. Journal homepage: http://ijece.iaescore.com Int J Elec & Comp Eng ISSN: 2088-8708  1541 Missing data, as indicated by the absence of data items for a subject, can obscure some potentially significant information. In practice, missing data has emerged as a significant determinant of data quality. Thus, the imputation of the missing value is needed [15]. Missing data is a common weakness in the classification problem, and it can cause the prediction system’s output to be ineffective [16], [17]. Ignoring missing data has an effect on the analysis’s results [18], [19], the outcomes of learning, as well as the outcomes of predictions on the collaborative prediction problem [20]. In quantitative studies, missing data leads to biased parameter estimates [21]–[24]. In the predictive model, the selection of methods for handling incorrect data missing can affect model performance [22], [25]. Missing data are common in medical research, and if not handled properly, they can result in a loss of statistical power and potentially biased results [26]–[28]. The standard data collection problems may involve noiseless data. In addition to the presence of noisy data, organizations face challenges with the presence of missing data. Missing data will affect extensive data collection, so investigating different filtering techniques for large data environments will be extraordinary [29]. This proposed study will not discuss or observe for the cause of the problem, rather than how to improve or predict the incorrect data with a technique which is commonly used in the case of missing data. This is the urgency of the research proposed to provide a reliable telemonitoring system with the smallest possible error rate to avoid misinterpretation. Several methods have been applied to predict missing data in various applications. In general, missing value imputation techniques fall into two categories: Statistical and machine learning-based techniques [30]. Expectation-maximization (EM), linear regression (LR), least squares (LS), and mean/mode are the four statistical techniques that are most frequently used [31]. The use of EM in the imputation of missing data has several advantages including missing data does not need to be ignored so that it can increase information for the accuracy of diagnosis [32] and can handle many patterns of missing data [33]. Imputation using linear regression results in a small standard deviation [34], although regression imputation is better than average imputation but results in biased parameter estimates [35]. Expert methods such as support vector machines (SVM) and artificial neural networks (ANN) used in data prediction were also reported in the study [36], [37]. However, this method has high computational costs and is complex to be implemented in computers with low memory resources. The literature study above provides enough knowledge as a basis for the proposed study. Research on predictions of missing data using a mathematical approach provides enough evidence to be applied to solve problems with ZigBee-based multiuser monitoring implementation. In this study, we applied a method to overcome the incorrect data, they are EM and regression linear imputation. This study aims to predict the incorrect data and determine the best method between the two proposed methods. Performance analysis is done by calculating the root mean square error (RMSE) (...truncated)