Forecast of Dengue Cases in 20 Chinese Cities Based on the Deep Learning Method

Abstract

Dengue fever (DF) is one of the most rapidly spreading diseases in the world, and accurate forecasts of dengue in a timely manner might help local government implement effective control measures. To obtain the accurate forecasting of DF cases, it is crucial to model the long-term dependency in time series data, which is difficult for a typical machine learning method. This study aimed to develop a timely accurate forecasting model of dengue based on long short-term memory (LSTM) recurrent neural networks while only considering monthly dengue cases and climate factors. The performance of LSTM models was compared with the other previously published models when predicting DF cases one month into the future. Our results showed that the LSTM model reduced the average the root mean squared error (RMSE) of the predictions by 12.99% to 24.91% and reduced the average RMSE of the predictions in the outbreak period by 15.09% to 26.82% as compared with other candidate models. The LSTM model achieved superior performance in predicting dengue cases as compared with other previously published forecasting models. Moreover, transfer learning (TL) can improve the generalization ability of the model in areas with fewer dengue incidences. The findings provide a more precise forecasting dengue model and could be used for other dengue-like infectious diseases.

Keywords: deep learning; dengue fever; forecast model; long short-term memory; transfer learning.

Figure 1

Spatial and temporal distribution of dengue cases in 20 selected cities in mainland China from 2005 to 2018. (A) Distribution of dengue cases in China (case numbers are distinguished by color and size according to the magnitude in each city), (B) the proportion of cases in each city, (C) time series of dengue incidence in mainland China (on the logarithmic scale), and (D–W) time series of dengue cases in the top 20 cities with the highest dengue incidence (on the logarithmic scale). Based on the Chinese provincial administrative districts public map downloaded from the National Geomatics Center of China (NGCC) website, this figure was produced using the matplotlib basemap toolkit (https://matplotlib.org/basemap/), which is a library for plotting data on maps in Python.

Figure 2

Summarized workflow for the construction of the LSTM-based forecasting model for dengue cases and its comparison with other candidate models. NNDSS: National Notifiable Disease Surveillance System; NMIC: National Meteorological Information Center; BPNN: Back Propagation Neural Network; GAM: Generalized Additive Model; SVR: Support Vector Regression; GBM: Gradient Boosting Machine.

Figure 3

The architecture of the dengue forecast model using the LSTM network.

Figure 4

Prediction dengue cases in the last 24 months by the long short-term memory (LSTM) model, back propagation neural network (BPNN) model, gradient boosting machine (GBM) model, generalized additive (GAM) model, and support vector regression (SVR) model. Comparison of 24-month predictions for 2017 to 2018 in Guangzhou, Foshan, Sipsong Panna, Dehong, and Chaozhou which pose a high degree of dengue infection.