TIME SERIES OUTLIER ANALYSIS FOR MODEL, DATA AND HUMAN-INDUCED RISKS IN COVID-19 SYMPTOMS DETECTION

Abstract

Information systems are important references aiming to support the decisions of decision-makers. Information reliability depends on the accuracy and efficacy of data and models. Therefore, some risks may emerge in information systems concerning models, data and humans. It is important to identify and extract outliers in decision support systems developed for the health information systems such as the detection system of Covid-19 symptoms. In this study, the risks that are important in decision making in Covid-19 symptom detection were determined by the statistical time series (ARMA) approach. Potential solutions are proposed in this way. Moreover, outliers are detected by software developed by using the Box-Jenkins model and reliability and accuracy of data is increased by using estimated data instead of outliers. In the implementation of this study, time-series-based data obtained from laboratory examinations of Covid-19 test devices can be used. With the method revealed here, outliers originating from healthcare workers or test apparatus can be detected and more accurate results can be obtained by replacing these outliers with estimated values.