Research on Gas Concentration Prediction Based on the ARIMA-LSTM Combination Model

The current single gas prediction model is not sufficient for identifying and processing all the characteristics of mine concentration time series data. This paper proposes an ARIMA-LSTM combined forecasting based on autoregressive integrated moving average (ARIMA) long short-term memory (LSTM) recurrent neural network. In model, ARIMA used to process historical data obtain corresponding linear results residual series. LSTM in further analysis series, predicting nonlinear factors are compared separately with those two models. Finally, RMSE, MAPE R2 evaluate accuracy three study show that metrics = 0.9825, 0.0124 RMSE 0.083. has highest lowest error more suitable predictive By comparing a data, applicability, validity scientificity proposed this verified, which great importance accurate early warning underground danger coal mines.