Prediction Model Of Air Quality Index Based On Optimized Deep Belief Network

In recent years,with the rapid development of our country's economy and energy consumption,haze weather is represented by the air pollution is increasingly serious,in our country present a spread to a wide range,high frequency and last for a long time.Due to its enormous adverse impact on social production and public health and the direct or indirect economic loss,it has received wide attention from all sectors of society.Air pollution problems of governance is a long-term process,in the short term can't under the premise of greatly improving air quality,it is necessary to the forecast of air quality,better reflect the change trend of air quality,so that government departments and the public in a timely manner to take protective measures,to prevent serious pollution incident.In this paper,the prediction of air quality index is carried out by measuring the pollution level of air pollution.AQI data and meteorological data in Beijing area depth based on belief network forecast model,by using the model to forecast the air quality index,the prediction accuracy is 91.4%,average sliding relatively integrated autoregressive model and BP neural network model is improved by 32.2% and 25.7% respectively,the predicted results reflect the deep learning model in air quality index prediction problem than traditional neural network model and statistical model has better prediction ability,under the background of big data can better play to the advantages of deep learning.Then in view of the deep belief network model in feature extraction part between the layers of connection weights and threshold of nodes random initialization problem,the introduction of differential evolution algorithm to optimize the model,and further improve the prediction precision,the results show that after differential evolution algorithm to optimize the depth of the belief network model in prediction accuracy of air quality index was 94.3%,increased by 2.9%,a deep belief network model shows that the optimization algorithm is adopted to deep learning model of optimization scheme is feasible,for deep learning model is optimized to improve model performance provides a train of thought.