Khan, Hafiz Aziz and Nadia, Nusrat Yasmin and Rabby, Habibor Rahman and Lindon, Abdur Rahman and Ferdus, Mst Zannatun and Arif, Md Habibul (2024) DATA-DRIVEN EPIDEMIOLOGICAL MODELING USING MACHINE LEARNING FOR DISEASE SPREAD FORECASTING AND PUBLIC HEALTH DECISION SUPPORT IN THE UNITED STATES. International Journal of Applied Mathematics, 37 (6). pp. 178-192. ISSN 1314-8060

Text paper+1+2024.pdf - Published Version Download (467kB)

Abstract

The issue of proper prediction of the spread of infectious diseases is one of the most important tasks of the state health organizations in the United States, especially during the outbreak of an epidemic that rapidly develops. The SusceptibleInfectedRecovered (SIR) framework is a traditional epidemiological framework that is based on the system of differential equations and relies on such parameters as the transmission rate ( β ) and the recovery rate ( γ ). Although these models offer useful theoretical frameworks, their use of fixed assumptions constrains their capacity to reflect complicated, nonlinear, and time-varying dynamics that are seen in outbreaks in the real world. This paper introduces an epidemiological modelling framework based on machine learning, which combines classical mathematical modelling with the use of machine learning to enhance disease spread prediction and aid in making decisions related to the health of the population. With time-series epidemiological data, prognostic models are then built in an attempt to approximate nonlinear functions that can be expressed as 9 y = f(X; 7), where X represents multidimensional input quantities and 7 represents model parameters that are estimated by optimization. It uses advanced algorithms, such as the Long Short-Term Memory (LSTM) networks and ensemble learning approaches, to represent temporal dynamics and stochastic change in the transmission patterns. Quantitative measures of model performance, like Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), are used to assess model performance. Findings show that neural network-enhanced models can greatly boost the predictive accuracy and flexibility over conventional methods. Such results highlight the importance of incorporating mathematical epidemiology with data-driven approaches to improve early warning, resource allocation, and inform evidence-based public health actions.

Item Type:	Article
Subjects:	A General Works > AI Indexes (General)
Depositing User:	admin eprints
Date Deposited:	07 Jun 2026 23:26
Last Modified:	07 Jun 2026 23:26
URI:	http://eprints.umsida.ac.id/id/eprint/16533

Actions (login required)

View Item