|
Title:
|
DEEP LEARNING FORECASTS OF HEPATITIS A, B, AND C INCIDENCE IN WAR-TORN UKRAINE: A MULTIVARIATE LSTM APPROACH |
|
Author(s):
|
Dmytro Chumachenko, Mykola Butkevych, Tetyana Chumachenko, Alexander Kirpich, Lyudmyla Kirichenko and Olha Matsyi |
|
ISBN:
|
978-989-8704-71-9 |
|
Editors:
|
Paula Miranda and Pedro Isaías |
|
Year:
|
2025 |
|
Edition:
|
Single |
|
Keywords:
|
Epidemic Model, Hepatitis A, Hepatitis B, Hepatitis C, LSTM, Deep Learning, Machine Learning |
|
Type:
|
Poster |
|
First Page:
|
321 |
|
Last Page:
|
323 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
if you are a member please login 
|
|
Paper Abstract:
|
Viral hepatitis remains a major public health threat globally, yet Ukraine's surveillance capacity has been repeatedly
destabilised by the Russian war, complicating resource planning for hepatitis A, B and C. We extracted 84 monthly case
counts for each virus (January 2018 - December 2024) from the Public Health Centre of Ukraine registry and trained a
two-layer Long Short-Term Memory network with a 12-month input window to generate 12-step forecasts. The model,
optimised with Adam and early stopping, achieved mean-absolute-percentage errors of 5.40 % (acute HAV), 13.78 %
(acute HBV), 17.13 % (chronic HBV), 26.34 % (acute HCV) and 23.89 % (chronic HCV) on the held-out 2024 horizon.
Critically, cumulative MAPE remained below 1.5 % for all series. These findings demonstrate that a lightweight,
CPU-deployable LSTM can deliver procurement-grade, year-ahead incidence forecasts for Ukraine's three principal
hepatitides, offering a scalable decision-support tool aligned with the WHO 2030 elimination agenda. |
|
|
|
|
|
|
