Scholarlink

Dengue fever is a disease transmitted by the Aedes aegypti mosquito and has been a public health concern in tropical and subtropical regions. Its incidence is influenced by climatic factors, such as temperature and precipitation, which favor the development of the vector. This study aims to predict the incidence of dengue cases in the city of Londrina, Paraná, between 2013 and 2024, using predictive models based on time series analysis. Data on weekly dengue cases were obtained from the InfoDengue system and combined with climatic variables, such as temperature, humidity, and precipitation. The methodology includes the application of the time series model called Seasonal Autoregressive Integrated Moving Average with exogenous variables (SARIMAX), which incorporates climatic variables in the predictive process. In addition, Long Short-Term Memory (LSTM) recurrent neural networks were used as a comparative, aiming to explore the ability of neural networks to capture complex temporal patterns. The SARIMAX (2,1,8)(1,1,1)52 model includes the covariates and obtained the best fit with the minimum temperature, presenting an excellent fit to the data and good forecasting. On the other hand, LSTM neural networks, despite their greater complexity and deep learning capacity, demonstrated great potential to capture the temporal and seasonal variations of the disease, although with greater volatility in the predictions compared to SARIMAX, representing a significant advance for the improvement of dengue predictions. This study suggests that the SARIMAX model, integrated with climate variables, is an effective tool for predicting the evolution of dengue cases and can serve as a basis for the management and planning of preventive actions. In addition, it reinforces the importance of improving LSTM neural networks to improve the accuracy of predictions in highly complex scenarios, with direct implications for the control and prevention of dengue.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting Using SARIMAX Model and LSTM Approach for Arboviruses Using Exogenous Climatic Variables in Predicting Dengue Incidence

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Claudia Stoeglehner Sahd,
  • Elisângela Ap. da Silva Lizzi,
  • Glaucia Maria Bressan,
  • Laurival Antonio Vilas-Boas

摘要

Dengue fever is a disease transmitted by the Aedes aegypti mosquito and has been a public health concern in tropical and subtropical regions. Its incidence is influenced by climatic factors, such as temperature and precipitation, which favor the development of the vector. This study aims to predict the incidence of dengue cases in the city of Londrina, Paraná, between 2013 and 2024, using predictive models based on time series analysis. Data on weekly dengue cases were obtained from the InfoDengue system and combined with climatic variables, such as temperature, humidity, and precipitation. The methodology includes the application of the time series model called Seasonal Autoregressive Integrated Moving Average with exogenous variables (SARIMAX), which incorporates climatic variables in the predictive process. In addition, Long Short-Term Memory (LSTM) recurrent neural networks were used as a comparative, aiming to explore the ability of neural networks to capture complex temporal patterns. The SARIMAX (2,1,8)(1,1,1)52 model includes the covariates and obtained the best fit with the minimum temperature, presenting an excellent fit to the data and good forecasting. On the other hand, LSTM neural networks, despite their greater complexity and deep learning capacity, demonstrated great potential to capture the temporal and seasonal variations of the disease, although with greater volatility in the predictions compared to SARIMAX, representing a significant advance for the improvement of dengue predictions. This study suggests that the SARIMAX model, integrated with climate variables, is an effective tool for predicting the evolution of dengue cases and can serve as a basis for the management and planning of preventive actions. In addition, it reinforces the importance of improving LSTM neural networks to improve the accuracy of predictions in highly complex scenarios, with direct implications for the control and prevention of dengue.