Research And Implementation Of Air Pollution Indicator Prediction Method Based On Deep Belief Network

At present,the methods of predicting air pollution indicator still have the problem of low accuracy,which leads to the less applicable value of the results of predicting air pollution indicator.At the same time,the current systems of predicting air pollution indicator also have the problem of poor interaction,which cannot make the results of predicting air pollution indicator been easily obtained by people.Therefore,this thesis has carried out research on the accuracy and interactivity of predicting air pollution indicator.The main work includes:(1)Aiming at the problem of the less accurate prediction of the current air pollution indicator,a prediction method of air pollution indicator based on deep belief network is proposed.The method uses the algorithm of canonical correlation analysis to analyze the correlate degree of the input data,and removes the pollution indicators and the meteorological indicators having low correlation with the indicator to be predicted.Moreover,the air pollution index data of C hours before t hour is used as the input of the deep belief network model,and the predicted value of t+1 hour is output.The network model is updated according to the error between the predicted value and the actual value.And the results of predicting air pollution indicator are compared under different prediction windows,accordingly,an optimal window of prediction is determined,which further improves the accuracy of prediction.(2)Aiming at the problem of the poor interaction mode of predicting air pollution indicator,an air pollution indicator prediction system based on multiple interaction modes is designed and implemented.The system includes a data acquisition subsystem,a data management subsystem,and a data processing subsystem.In the data acquisition subsystem,the air pollution indicator data is collected in real time using the terminal monitoring device.In the data management subsystem,the cloud platform is used to implement access management of monitoring data.In the data processing subsystem,the online prediction method of air pollution indicator based on deep belief network is used to analyze and predict the collected data,and various interaction methods are adopted for users to obtain data and the results of predicting air pollution indicator.The interaction mode includes cloud interaction of data,mobile terminal interaction of data and three-dimensional visual interaction of data,so that users can not only query the concentration value of air pollution indicator more conveniently,but also receive the alarm and prediction information of air pollution indicator exceeding the standard in time.