The Research Of Airline Passenger Flow Prediction Based On Wavelet Neural Network

In recent years, many airlines are focus on airline passenger flow as it grows rapidly. So we need to build appropriate and high accuracy airline passenger prediction model for airlines. The right prediction approach not only brings profit but also offers help to make a decision.In order to forecast airline passenger flow, firstly, we need get all airlines daily passenger flow from massive travel record of passenger in recent years. Secondly, wavelet transform algorithm is used to flow data de-noising which has reduced noise produced by accident and hold its intrinsic characteristics.Wavelet transform decomposes complex signals into several detailed parts so that we can observe fluctuation in detail. The prediction accuracy of BP neural network will be improved when it combines with wavelet transform. We need make a distinction between weekday flow and holiday flow because of their dramatic difference flow fluctuation. The flow waves among near weekday are very similar through analyzing flow data. In order to improve forecast precision, we divide weekday flow into seven groups by different week types. Meanwhile, this paper structures BP neural network and embedded wavelet neural network for the nonlinear workweek flow forecast respectively. For the holiday prediction, the article also builds two forecasting methods including coefficient of variation and loose wavelet neural network with similar samples. Both of them are based on the similar wave curves during the same holiday of past few years.Finally the paper takes experiments with Beijing to Guangzhou route passenger flow in MATLAB. Results show that the embedded wavelet neural network approach is a befitting and appropriate method for weekday passenger flow prediction and the loose wavelet neural network with similar samples approach is a feasible and high accuracy method for holiday passenger flow prediction.