Prognosis of Dengue Incidences in India Using LSTM with Multivariate Data

Monitoring climatic change and its impacts unique to the disease is required for predicting the expansion of dengue. One of the infectious diseases transmitted by vectors with the fastest rate of transmission is dengue. In this study, we have proposed a forecasting model for the prediction of dengue cases by considering the climatic variables for all the states of India for last 12 years. For our study, we have accumulated data with factors like temperature (minimum and maximum), relative humidity, mean wind speed, dengue cases and deaths. For the prediction, the different models ANN, RNN, LSTM, Stacked LSTM, BiLSTM for forecasting time series prediction are implemented and then compared. During our study, it is found that humidity and maximum temperature have a strong correlation with the number of dengue cases. BiLSTM model outperformed all other deep learning model with an accuracy of 98.91%. This study will assist the healthcare practitioners to take informed decisions beforehand.