Data mining for air traffic flow forecasting: a hybrid model of neural network and statistical analysis

Abstract

The objective of this paper is to build an air traffic flow prediction model. The air traffic flow prediction plays a key role in the airspace simulation model and air traffic flow management system. In China the air traffic information in each regional control center has not integrated together by now. The information only in a single regional control center can not reach the requirement of the current method based on 4-dimensional trajectory prediction. The new method is needed to solve this problem. Large collection of radar data is stored. But there is no effort made to extract useful information from the database to help in the estimation. Data mining is the process of extracting patterns as well as predicting previously unknown trends from large quantities of data. Neural network and statistics are frequently applied to data mining with various objectives. This paper employs neural networks combined with the statistical analysis of historical data to forecast the traffic flow. Two models with different types and input data are proposed. The accuracy of two models is tested and compared to each other using flow data at an arrival fix in Beijing control center. The result shows that these models are feasible for practical implementations. The suitable models for different prediction conditions are also suggested.