Optimization of Urban Air Pollutant ConcentrationPrediction and Health Risk Assessment Based onLSTM Model in Healthy Urban Space:A Case Study of Changsha-Zhuzhou-XiangtanUrban Agglomerations

With the improvement of people's living standards, more people are concerned about the air quality and safety of residential cities, and the concept of healthy urban space is gradually becoming deeply rooted in people's hearts.This study is based on long and short term memory neural network algorithms, incorporating AMs into them.The research adjusts the data input to the algorithm according to spatiotemporal characteristics and incorporates a stack-type self-coding network into an improved long and short term memory neural network to predict the concentration of urban air pollutants.The air pollutant data of Changsha-Zhuzhou-Xiangtan is used to test the model, and the test results are as follows: The index values of the mean absolute error and coefficient of determination of the intelligent prediction model with all improvement measures in the test set are 4.0 and 0.94, respectively, which is significantly better than the traditional and partially improved long and short term memory neural network.The algorithm model with complete improvement measures is selected for comparative experiments with other recurrent neural networks.This experimental result shows that the overall fluctuation amplitude of this model is the smallest under various test sample numbers.The mean absolute error and root-mean-square error on the whole test set are 6.7 and 9.2, respectively, which are still higher than other models.At this time, the memory consumption is 81 MB, 117 MB, and 154 MB, and the memory consumption is also lower.The experimental data proves that this model, combined