The Research On Flight Delays Prediction Model And Method Based On Data Mining

Flight delays have plagued the international and domestic civil aviation industry is a hot issue. In recent years China Airlines is increasing the delay has affected the aviation industry, to improve the situation of imminent delays. For the busy hub airports, the delays in warning the number of flights the airport is of great significance. This is because the number of delayed flights means that the number of stranded passengers, when a certain number of stranded passengers (or a certain number of delayed flights), the airport must take the appropriate plan for intervention. Therefore, if advance warning of the number of delayed flights, the airport is undoubtedly related departments will be very helpful. This article on a busy airport flight information table as the main object of study, for each time period in which the number of delayed flights modeling and forecasting.As society and the advancement of technology, people can gather to become larger and larger amount of data, these data processing and therefore difficult to predict become very large. The emerging one by one dealing with massive data mining technology to solve this problem prediction provides a new opportunity. Prediction is an important data mining research, data mining, pattern recognition, statistical data analysis and other fields have a wide range of applications. This flight delays for time series prediction model was established, the use of data mining in the prediction of several common methods such as linear regression, nonlinear regression, neural networks, applied to the data on flight delays, and achieved good prediction results. These prediction methods and some of the existing prediction algorithms are usually based on the data, but in practice many do not have prior knowledge to the full use of, for example, major holidays and special dates and so on. This paper presents a fusion of a priori knowledge of the support vector machine regression. Defined by the prior knowledge constraints on the data set, the model of learning and training. Because of this new method to join the prior knowledge, which some of the existing method of prediction than the effect more visible. This paper also experiments with the standard support vector machine regression method were compared with experimental results show the effectiveness of the new forecasting method.