The Application of Time Series Models in the Prediction of Hand-Foot-Mouth Disease Incidence

Abstract

Hand-Foot-Mouth Disease (HFMD) is a contagious illness predominantly affecting infants and children under five years old, caused by human enteroviruses. Over the past five decades, HFMD has rapidly spread across the Asia-Pacific region, gradually evolving into a significant public health challenge for many countries within this area. Currently, HFMD has emerged as an increasingly severe public health issue in our country. Therefore, analyzing the influencing factors of HFMD and predicting its incidence trends are of paramount importance for the prevention of the disease. With the rapid advancement of artificial intelligence technology, predictive models employing deep learning techniques have demonstrated superior performance among various infectious disease prediction models. This paper aims to construct a predictive model using deep learning methods to further enhance the accuracy of HFMD incidence predictions. We compared the effectiveness of Long Short-Term Memory (LSTM) networks, Transformer, and Informer models in HFMD prediction. The research findings indicate that the Informer model, by utilizing self-attention mechanisms and convolutional neural networks, can more effectively address long-term dependencies in time series data, thereby showing better performance in HFMD prediction compared to the LSTM and Transformer models. This has led to improvements in prediction accuracy and generalization capability.