A hybrid ARIMA-LSTM model optimized by BP in the forecast of outpatient visits

Effective hospital outpatient forecasting is an important prerequisite for modern hospitals to implement intelligent management of medical resources. As outpatient visits flow may be complex and diverse volatility, we propose a hybrid Autoregressive Integrated Moving Average (ARIMA)-Long Short Term Memory (LSTM) model, which hybridizes the ARIMA model and LSTM model to obtain the linear tendency and nonlinear tendency correspondingly. Instead of the traditional methods that artificially assume the linear components and nonlinear components should be linearly added, we propose employing backpropagation neural networks (BP) to imitate the real relationship between them. The proposed hybrid model is applied to real data analysis and experimental analysis to justify its performance against single ARIMA model, single LSTM model and the hybrid ARIMA-LSTM model based on the traditional method. Compared with competitors, the proposed hybrid model produced the lowest RMSE, MAE and MAPE. It achieves more accurate and stable prediction. Therefore, the proposed model can be a promising alternative in outpatient visit predictive problems.