Application of a long short-term memory neural network: a burgeoning method of deep learning in forecasting HIV incidence in Guangxi, China.

摘要: Guangxi, a province in southwestern China, has the second highest reported number of HIV/AIDS cases China. This study aimed to develop an accurate and effective model describe tendency HIV predict its incidence Guangxi. data Guangxi from 2005 2016 were obtained database Chinese Center for Disease Control Prevention. Long short-term memory (LSTM) neural network models, autoregressive integrated moving average (ARIMA) generalised regression (GRNN) models exponential smoothing (ES) used fit data. Data 2015 validate most suitable models. The performances evaluated by evaluating metrics, including mean square error (MSE), root error, absolute percentage error. LSTM had lowest MSE when N value (time step) was 12. appropriate ARIMA (1, 1, 2) (0, 2)12 (2, 0) 2)12, respectively. accuracy GRNN ES forecasting relatively poor. Four performance metrics all lower than ARIMA, more other time-series is important monitoring control local epidemics.