Time Series Forecasting Fusion Network Model Based on Prophet and營mproved LSTM

Time series forecasting and analysis are widely used in many fields application scenarios. historical data reflects the change pattern trend, which can serve decision each scenario to a certain extent. In this paper, we select time prediction problem atmospheric environment start research. terms of support, obtain nearly 3500 vehicles some cities China from Runwoda Research Institute, focusing on major pollutant emission non-road mobile machinery high Beijing Bozhou, Anhui Province build dataset conduct experiments them. This paper proposes P-gLSTNet model, uses Autoregressive Integrated Moving Average model (ARIMA), long short-term memory (LSTM), Prophet predict compare emissions future period. The validated four public sets one self-collected set, mean absolute error (MAE), root square (RMSE), percentage (MAPE) selected as evaluation metrics. experimental results show that proposed fusion predicts less error, outperforms backbone method, is more suitable for time-series scenario.