Analysis And Prediction Of LSTM Based On TensorFlow In Taiyuan Air Quality AQI Index

Atmospheric pollution refers to people’s production activities,because some inappropriate behaviors discharge pollutants into the atmosphere,pollutants discharged into the atmosphere reach a certain concentration will cause harm to people’s bodies,while It will also affect the development of other species.Contaminants are usually changing and dynamic,so the factors that cause air pollution are also diverse.For example,we know that sulfur dioxide,nitrogen dioxide,and PM2.5 are the main culprit in air pollution.At the same time,some weather factors will also affect the air pollution,such as: precipitation,wind power,wind direction and so on.Shanxi Taiyuan is a cultural ancient city with a history of 2,500 years.For a long time,Shanxi Province,as a major coal province in the country,has contributed to the development of the country,and it has inevitably caused certain environmental impacts.Nowadays,environmental problems are becoming increasingly strict.Heavy,especially during the winter heating period,Shanxi mainly relies on coal-fired heating,and the environmental problems are very serious.Shanxi Province wants to restore the blue sky and blue sky,especially as the capital city of Taiyuan.It is even more necessary to meet the challenges and developments of the future with a new look.After many years of rectification in Taiyuan City,the environmental improvement has not achieved the expected results.In order to further reveal and control the pollution of Taiyuan air quality,it is necessary to understand the trend of air change,and to grasp timely,accurate and comprehensive air quality information,air quality AQI.Accurate prediction of the index is one of the effective measures.Since the air quality AQI index is affected by multiple hard-to-determine and non-linear influence factors,traditional prediction methods lead to low prediction accuracy and low efficiency.Among multiple neural network models,based on long-short-term memory units(Long Short-term)The recurrent neural network model of memory,LSTM can effectively predict the information-dependent ability of long-short clustering in time series data,and can effectively predict the air quality AQI index.This paper firstly separates the correlation between various pollutants and the AQI index,and based on Python 3.5.2 and TensorFlow,combined with the trend of various influencing factors of air quality in recent years,the AQI index of air quality is predicted,using the mean square error.(MSE)performs error analysis on the predicted data and the original data,and concludes that the TensorFlow-based LSTM neural network can accurately predict the air quality AQI index.The article first uses Ri386 3.3.3 to analyze the correlation between air pollutants and AQI index,and introduces the influence factors of weather factors on air quality,and analyzes the influencing factors of Taiyuan air quality AQI index from various aspects.Then based on Python 3.5.2 and TensorFlow framework,combined with the trend of various influencing factors of air quality in recent years,predict the AQI index of Taiyuan air quality,and finally use the mean square error(MSE)to make errors on the predicted data and the original data.analysis.The main research contents are as follows:(1)Analysis using Ri386 3.3.3 Changes in air quality in Taiyuan air quality in different years,and related factors affecting the AQI index,including various pollutants and climatic conditions.(2)The air quality AQI index is predicted by SVM,ARIMA time series and LSTM,respectively,and then the average error and relative error curve are compared.The experimental results show that the LSTM based on Tensorflow is superior to the ARIMA time series model and SVM in terms of prediction accuracy,error rate and reliability.