On Comparing Cross-Validated Forecasting Models with a Novel Fuzzy-TOPSIS Metric: A COVID-19 Case Study

Sustainability, Journal Year: 2021, Volume and Issue: 13(24), P. 13599 - 13599

Published: Dec. 9, 2021

Time series cross-validation is a technique to select forecasting models. Despite the sophistication of over single test/training splits, traditional and independent metrics, such as Mean Absolute Error (MAE) Root Square (RMSE), are commonly used assess model’s accuracy. However, what if decision-makers have different models fitting expectations each moment time series? What precision forecasted values also important? This case predicting COVID-19 in Amapá, Brazilian state Amazon rainforest. Due lack hospital capacities, model that promptly precisely responds notable ups downs number cases may be more desired than average only good performances frequent calm circumstances. In line with this, this paper proposes hybridization Technique for Order Preference by Similarity Ideal Solution (TOPSIS) fuzzy sets create similarity metric, closeness coefficient (CC), enables relative comparisons under heterogeneous considers volatility predictions. We present study using three parametric machine learning forecast numbers. The results indicate introduced metric informative performance assessment especially when cross-validation.

Language: Английский

---

Learning and Predicting from Dynamic Models for COVID-19 Patient Monitoring

Statistical Science, Journal Year: 2022, Volume and Issue: 37(2)

Published: May 1, 2022

COVID-19 has challenged health systems to learn how learn. This paper describes the context, methods and challenges for learning improve care at one academic center. Challenges include: (1) choosing a right clinical target; (2) designing accurate predictions by borrowing strength from prior patients’ experiences; (3) communicating methodology clinicians so they understand trust it; (4) patient moment of decision; (5) continuously evaluating revising adapt changing patients demands. To illustrate these challenges, this contrasts two statistical modeling approaches—prospective longitudinal models in common use retrospective analogues complementary context—for predicting future biomarker trajectories major events. The are applied validated on cohort 1678 who were hospitalized with during early months pandemic. We emphasize graphical tools promote physician inform decision making.

Language: Английский

---

Review of Machine Learning-Based Disease Diagnosis and Severity Estimation of COVID-19

Advances in intelligent systems and computing, Journal Year: 2023, Volume and Issue: unknown, P. 1 - 18

Published: Jan. 1, 2023

Language: Английский

---