Research On Environmental Pollution Time Series Prediction Model Based On LSTM

With the development of industry and the popularization of automobiles,the problem of air pollution is becoming more and more serious.How to accurately predict the composition of air pollution is of great significance to the treatment of environmental pollution.The traditional time series prediction method has the shortcomings of large error and long time.With the development of deep learning,the air pollution components can be accurately analyzed by using neural network method.Taking environmental pollution as the research object,it is of profound research significance and practical application value to study an efficient time series data prediction method based on neural network.A more efficient long-term memory neural network is designed.The logic architecture of neural network is optimized,and the relevant parameters are selected by grid search algorithm.Explore the influence of different parameters on the prediction results.Parameters such as activation function,loss function and neuron number were selected to reduce errors and time consumption.The newly constructed neural network model was compared with neural networks such as LSTM and GRU to observe whether the error and time consumption were reduced.Verify whether the multi-layer network model is better than the single-layer network model,and evaluate the prediction error.The improved prediction model uses the stacked model,selects the LSTM layers according to the characteristics of the data,and optimizes the model by adjusting each parameter.In addition,this model mainly forecasts the time series data.Taking environmental pollution data as an example,the experimental results show that compared with the RNN,LSTM,GRU and other neural networks,the improved neural network has lower error rate and shorter time,which is more suitable for the prediction of environmental pollution time series data.At the same time,other sample data are used to test the neural network to verify its universal applicability.