Study On The Application Of BP Neural Network In Air Quality Prediction Based On Improved Fruit Fly Optimization Algorithm

In recent years,the continuous development and improvement of China's economy has brought with the problem of increasing energy consumption and environmental pollution."Haze" and air pollution have also become a hot topic in people's daily life.How to effectively use the existing historical data of various cities to analyze and predict the urban air quality more accurately and effectively can not only bring convenience to people's life and travel,but also provide certain help for the treatment of air pollution.The main content of this paper is to take the air quality index of Nanchang city as the research object and establish a prediction model for air quality of Nanchang city based on BP neural network.Firstly,based on the improvements proposed by the Fruit Fly Optimization Algorithm and other related scholars,the algorithm is further optimized,and an improved ACFOA algorithm is proposed.The improvement mainly includes adding a jump parameter to the judgment value of the taste concentration of the fruit fly,which improves the defect that the value cannot be less than zero,and uses the Tent mapping instead of the Logistic mapping,making the chaotic mapping The uniform distribution improves the optimization ability of the algorithm.In addition,experiments were performed to prove the performance of the improved algorithm.Finally,the improved algorithm is used as the algorithm of initialization weight and threshold of BP neural network.Next,by analyzing the calculation method of the air quality index AQI,two types of predicted air quality indexes are proposed.One is to use the AQI value and the weather data of the day to predict the AQI value of the next day.The other is to predict the IAQI of each pollutant according to the content of each pollutant in the air and the current weather data first,and then use the predicted IAQI of each pollutant to calculate AQI value collectively.The results show that the improved ACFOA-BP algorithm has improved performance compared with BP algorithm and FOA-BP algorithm,and the consumed time is not much different.The performance of using the indirect prediction method is slightly better than the direct prediction method,but the indirect forecast takes about 6-7 times longer than the direct forecast.