Airport Noise Heterogeneous Ensemble Prediction Model Based On Observational Learning

With the development of China’s overall national strength, civil aviation industry has gotten great achievement. But the airport noise pollution becomes more and more serious.How to control airport noise is very important for civil aviation personnel. Airport noise prediction is an important prerequisite for airport environment assessment and noise control work, therefore, it means a lot to make a scientific, reasonable and comprehensive airport noise prediction model.The existing airport noise prediction methods based on machine learning are analyzed.Most of incidence analysis methods only have single learner, so that their effects are not good.This paper presents an ensemble prediction method of airport noise incidence analysis. In this paper, the major noise influential factors are analyzed. And all base learners, which are built based on space fitting and BP neural network, are combined by observational learning algorithm. The prediction accuracy is improved effectively by integrating multiple heterogeneous airport noise prediction base learners.The airport noise time series prediction method based on Kalman filtering optimizing is presented by optimizing the prediction result. This method firstly builds time series by using noise statistics, then trains prediction learner based on support vector regression, finally optimizes the prediction results using Kalman filtering. Because of the reasonable times series structure method and the optimization for the result, this method is better than the former.Finally, in order to fit the prediction demands such as stable, accurate and reliable, an airport noise heterogeneous ensemble prediction model based on observational learning algorithm is proposed in this paper. This model unifies uses the training data set, and then combines association analysis prediction and time series prediction reasonably by opposite observational learning algorithm. The proposed airport noise heterogeneous ensemble prediction model has higher accuracy and stronger generalization ability than that of the single learner prediction model, and can be put into use in most China airport.