Deep learning models for forecasting dengue fever based on climate data in Vietnam

Abstract

Background: Dengue fever (DF) represents a significant health burden in Vietnam, which is forecast to worsen under climate change. The development of an early-warning system for DF has been selected as a prioritised health adaptation measure to climate change in Vietnam.

Objective: This study aimed to develop an accurate DF prediction model in Vietnam using a wide range of meteorological factors as inputs to inform public health responses for outbreak prevention in the context of future climate change.

Methods: Convolutional neural network (CNN), Transformer, long short-term memory (LSTM), and attention-enhanced LSTM (LSTM-ATT) models were compared with traditional machine learning models on weather-based DF forecasting. Models were developed using lagged DF incidence and meteorological variables (measures of temperature, humidity, rainfall, evaporation, and sunshine hours) as inputs for 20 provinces throughout Vietnam. Data from 1997-2013 were used to train models, which were then evaluated using data from 2014-2016 by Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).

Results and discussion: LSTM-ATT displayed the highest performance, scoring average places of 1.60 for RMSE-based ranking and 1.95 for MAE-based ranking. Notably, it was able to forecast DF incidence better than LSTM in 13 or 14 out of 20 provinces for MAE or RMSE, respectively. Moreover, LSTM-ATT was able to accurately predict DF incidence and outbreak months up to 3 months ahead, though performance dropped slightly compared to short-term forecasts. To the best of our knowledge, this is the first time deep learning methods have been employed for the prediction of both long- and short-term DF incidence and outbreaks in Vietnam using unique, rich meteorological features.

Conclusion: This study demonstrates the usefulness of deep learning models for meteorological factor-based DF forecasting. LSTM-ATT should be further explored for mitigation strategies against DF and other climate-sensitive diseases in the coming years.

Fig 1. Yearly DF incident cases per 100,000 population (log-scaled) for 20 different provinces in northern, central, and southern Vietnam from 1997 to 2016.

In the box and whisker plots, green dots indicate mean values.

Fig 2. Data processing pipeline.

NIHE = National Institute of Hygiene and Epidemiology. IMHEN = Vietnam Institute of Meteorology, Hydrology and Climate Change. DF = dengue fever.

Fig 3. Prediction performances of CNN, LSTM, and LSTM-ATT during the last 36 months in six Vietnamese provinces.

Predicted incidence rates per 100,000 population from 2014 to 2016 are shown compared to the observed incidence rates. The closer the predictions are to the observed values, the better the prediction accuracies. CNN = convolutional neural network. LSTM = long short-term memory. LSTM-ATT = attention mechanism-enhanced LSTM.

Fig 4. RMSEs and MAEs for all models (LSTM, and LSTM-ATT, CNN, Transformer) for all 20 provinces.

The smaller the values, the better the prediction accuracies. RMSE = root mean square error. MAE = mean absolute error. CNN = convolutional neural network. LSTM = long short-term memory. LSTM-ATT = attention mechanism-enhanced LSTM.

Fig 5. DF forecasting models with RMSE- and MAE-based rankings.

Rankings are based on the relative scores for lowest RMSE or MAE in the prediction of dengue fever one month ahead. Grey-outlined circles indicate mean values. RMSE = root mean square error. MAE = mean absolute error. LSTM = long short-term memory. LSTM-ATT = attention mechanism-enhanced LSTM. CNN = convolutional neural network. Poisson = Poisson regression. XGB = XGBoost Extreme Gradient Boosting. SVR = Support Vector Regressor with Radial Basis Kernel. SVR-L = Support Vector Regressor with Linear Kernel. SARIMA = Seasonal Autoregressive Integrated Moving Average.

Fig 6. Outbreak detection by LSTM-ATT.

Numbers of actual outbreak months, correct outbreak month predictions (true positive) and incorrect outbreak month predictions (false positive) for each province are shown (Fig 6A). Additionally, prediction metrics (precision, accuracy, sensitivity, and specificity) for each province are displayed (Fig 6B). If a province did not have any actual epidemic months in the evaluation period, the precision and sensitivity are not available. LSTM-ATT = attention mechanism-enhanced LSTM.

Fig 7. Performance of multi-step ahead predictions of LSTM-ATT for all provinces.

Error metrics are displayed for all 20 provinces (Fig 7A for RMSE and middle for MAE) in addition to the predicted and observed incidence rates per 100,000 population in three provinces (Fig 7B). LSTM-ATT = attention mechanism-enhanced LSTM. RMSE = root mean square error. MAE = mean absolute error.

Fig 8. Precision, accuracy, sensitivity, and specificity for multi-step ahead epidemic prediction using LSTM-ATT.

LSTM-ATT = attention mechanism-enhanced long short-term memory.