Forecasting COVID-19 Cases Using Dynamic Time Warping and Incremental Machine Learning Methods

Aug 5, 2026

Upload by Finn

Forecasting future pandemics depends on building effective strategies before large-scale spread. This study reconstructs COVID-19 dynamics across 2020 and evaluates case prediction for the 50 countries with the highest case counts. State-of-the-art models are compared with online incremental machine learning approaches. Three training strategies are tested, including single-country, multi-country, and similarity-based transfer using time series similarity measures such as Euclidean distance and Dynamic Time Warping, with ARIMA as baseline. Results show strong error reductions from the proposed approach.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Problem context and motivation

Methodology

Time series similarity measures and DTW/ED

Incremental machine learning and compared models

Training strategies: single-country, multi-country, similarity-based transfer

Experiments and Baseline

Country selection and clustering for similar countries

ARIMA baseline

Results and Findings

RMSE reductions and key advantages

Implementation and Code Availability

Technological University Dublin ARROW@TU Dublin Articles 2023 Forecasting COVID-19 Cases Using Dynamic Time Warping and Incremental Machine Learning Methods Luis Miralles-Pechuán Technological University Dublin, Ireland, [luis.miralles@tudublin.ie](luis.miralles@tudublin.ie) Ankit Kumar University College Dublin, Ireland Andres L. Suare\-Cetrulo University College Dublin, Ireland Follow this and additional works at: [https://arrow.tudublin.ie/creaart](https://arrow.tudublin.ie/creaart) Part of the Computer Engineering Commons, and the Medicine and Health Sciences Commons Recommended Citation Miralles-Pechuán, Luis; Kumar, Ankit; and Suare\-Cetrulo, Andres L., "Forecasting COVID-19 Cases Using Dynamic Time Warping and Incremental Machine Learning Methods" (2023) . Articles. 173. [https://arrow.tudublin.ie/creaart/173](https://arrow.tudublin.ie/creaart/173) This Article is brought to you for free and open access by ARROW@TU Dublin. It has been accepted for inclusion in Articles by an authorized administrator of ARROW@TU Dublin. For more information, please contact [arrow.admin@tudublin.ie](arrow.admin@tudublin.ie), [aisling.coyne@tudublin.ie](aisling.coyne@tudublin.ie), [gerard.connolly@tudublin.ie](gerard.connolly@tudublin.ie), [vera.ki](vera.ki)[lshaw@tudublin.ie](lshaw@tudublin.ie). This work is licensed under a Creative Commons Attribution-Share Alike 4.0 International License. Funder: This research received no external funding Received: 20 April 2022 Revised: 15 December 2022 Accepted: 10 January 2023 DOI: 10.1111/exsy.13237 O R IG INA L ARTI CLE Forecasting COVID-19 cases using dynamic time warping and incremental machine learning methods Luis Miralles-Pechuán 1 | Ankit Kumar 2 | Andrés L. Suárez-Cetrulo 2 1School of Computer Science, Technological University Dublin, Dublin, Ireland 2Centre for Applied Data Analytics Research (CeADAR), University College Dublin, Dublin, Ireland Correspondence Luis Miralles-Pechuán, School of Computer Science, Technological University Dublin, Dublin, Ireland. Email: [luis.miralles@tudublin.ie](luis.miralles@tudublin.ie) Abstract The investment of time and resources for developing better strategies is key to dealing with future pandemics. In this work, we recreated the situation of COVID-19 across the year 2020, when the pandemic started spreading worldwide. We conducted experiments to predict the coronavirus cases for the 50 countries with the most cases during 2020 . We compared the performance of state-of-the-art machine learning algorithms, such as long-short-term memory networks, against that of online incremental machine learning algorithms. To find the best strategy, we performed experiments to test three different approaches. In the first approach (single-country), we trained each model using data only from the country we were predicting. In the second one (multiple-country), we trained a model using the data from the 50 countries, and we used that model to predict each of the 50 countries. In the third experiment, we first applied clustering to calculate the nine most similar countries to the country that we were predicting. We consider two countries to be similar if the differences between the curve that represents the COVID-19 time series are small. Todo so, we used time series similarity measures (TSSM) such as Euclidean Distance (ED) and Dynamic Time Warping (DTW) . TSSM return a real value that represents the distance between the points in two time series which can be interpreted as how similar they are. Then, we trained the models with the data from the nine more similar countries to the one that was predicted and the predicted one. We used the model ARIMA as a baseline for our results. Results show that the idea of using TSSM is avery effective approach. By using it with the ED, the obtained RMSE in the singlecountry and multiple-country approaches was reduced by 74.21% and 74.70%, respectively. And by using the DTW,

FAQs

What time period and set of countries are used for the COVID-19 forecasting experiments?

The study recreates COVID-19 behavior over 2020 and predicts cases for the 50 countries with the most cases during that year.

How do the single-country, multiple-country, and similarity-based approaches differ?

In the single-country approach, each model trains only on the target country’s data. In the multiple-country approach, one model is trained on data from all 50 countries and then used to predict each country. In the similarity-based experiment, clustering selects the nine most similar countries by time-series curve differences, and models are trained using those countries’ data.

What role do Dynamic Time Warping and Euclidean distance play in improving prediction accuracy?

They are used as time series similarity measures to quantify how similar countries’ COVID-19 time-series curves are. Using these measures to guide training leads to large RMSE reductions compared with the baseline results.

Aug 5, 2026

Upload by Finn