Civil-aviation suspicious-order feature extraction method based on hybrid-feature selection algorithm

Abstract

The invention discloses a civil-aviation suspicious-order feature extraction method based on a hybrid feature selection algorithm. The method includes the steps such as: preprocessing data; selectinginitial features; calculating initial-feature information gain values, and generating pre-selected feature subsets; utilizing a decision tree C4.5 algorithm to select an optimal candidate feature subset; and generating a subset through a sequence forward-floating-searching algorithm, and utilizing the decision tree C4.5 algorithm to evaluate the subset to obtain a finally selected feature subset.The method has the advantages that different evaluation criteria are taken at different searching stages, the method has both a fast-calculation characteristic of a Filter algorithm and a high-classification-precision advantage of a Wrapper algorithm through a Filter-Wrapper hybrid feature selection algorithm of information gains and sequence forward-floating-searching, performance of the algorithm is guaranteed, time complexity of the algorithm is also reduced, the method has a dynamic-searching characteristic of the feature subsets, possibility of occurrence of feature nesting is avoided, and a higher suspicious-order recognition rate can also be obtained.

Description

technical field [0001] The invention belongs to the technical field of data mining for increasing civil aviation revenue, and in particular relates to a feature extraction method for suspicious civil aviation orders based on a mixed feature selection algorithm. Background technique [0002] In the traditional civil aviation business process, passengers purchase air tickets are divided into two important stages, one is flight seat reservation, and the other is ticket issuance. In fact, at the stage of passenger booking, only the seat reservation after the real ticket is issued can bring value to the airline, so there are a large number of potential revenue loopholes (Revenue Leakage, RL) between seat reservation and actual sales. For example, with the rapid expansion of agents and distributors in recent years, some agents use various methods such as entering false ticket numbers in the Passenger Name Record (PNR), entering false passenger names, and making repeated reservatio...

AI Technical Summary

Problems solved by technology

For example, with the rapid expansion of agents and distributors in recent years, some agents use various methods such as entering false ticket numbers in the Passenger Name Record (PNR), entering false passenger names, and making repeated reservations. Striving for more profits has resulted in huge losses for the airlines, and the revenue managers of the airlines have spent a lot of time and manpower to check the reservation records on the flight, and at the same time take corresponding measures for the order problems found

Images