Scholarlink

<p>The urgent situation during the COVID-19 pandemic led to uncertainty in COVID-19 statistics, such as infection rate, Hospitalization Rate (HR), and Death Rate (DR). This paper addresses the challenge of causal forecasting of COVID-19 severity measures while considering the uncertainty utilizing the Adaptive Neuro-Fuzzy Inference System (ANFIS). The causal and time series ANFIS forecasters are developed through PYTHON coding in the case study of Iran using a daily dataset of 1010 entries collected from 2020 to 2022. The results demonstrate the superiority of causal ANFIS forecasters according to accuracy measures of the training, test, and overall data sets. The accuracy of 96.9% and 99.3% for causal ANFIS forecasters of HR and DR, respectively, demonstrates the considerable capability of the created forecasters to provide accurate forecasts of COVID-19 severity measures. In addition, causal ANFIS forecasters provide the opportunity for accuracy sensitivity analysis according to the one-factor-at-a-time exclusion of indicators. The corresponding results reveal that although the vaccination rate is a significant indicator, infection and hospitalization rates are the most important indicators affecting the accuracy of causal ANFIS forecasters of HR and DR, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A neuro-fuzzy causal approach for pandemic severity forecasting: COVID-19 case study

  • 期刊论文

  • 发表日期：

    2025年5月31日

查看全文
  • Ramin Omrani,
  • Arash Nemati

摘要

The urgent situation during the COVID-19 pandemic led to uncertainty in COVID-19 statistics, such as infection rate, Hospitalization Rate (HR), and Death Rate (DR). This paper addresses the challenge of causal forecasting of COVID-19 severity measures while considering the uncertainty utilizing the Adaptive Neuro-Fuzzy Inference System (ANFIS). The causal and time series ANFIS forecasters are developed through PYTHON coding in the case study of Iran using a daily dataset of 1010 entries collected from 2020 to 2022. The results demonstrate the superiority of causal ANFIS forecasters according to accuracy measures of the training, test, and overall data sets. The accuracy of 96.9% and 99.3% for causal ANFIS forecasters of HR and DR, respectively, demonstrates the considerable capability of the created forecasters to provide accurate forecasts of COVID-19 severity measures. In addition, causal ANFIS forecasters provide the opportunity for accuracy sensitivity analysis according to the one-factor-at-a-time exclusion of indicators. The corresponding results reveal that although the vaccination rate is a significant indicator, infection and hospitalization rates are the most important indicators affecting the accuracy of causal ANFIS forecasters of HR and DR, respectively.