Research On Data Mining Model Of Revenue Increase Rules Based On Civil Aviation Order Data

With the population of civil aviation passenger electronic tickets,but the problem has increased more,such as many people use the airline loopholes to take seats,resulting in a lot of seats are wasted.For these problems,this paper hopes to research on civil aviation passenger service data,and use reasonable and effective data mining method to find the civil aviation loopholes,providing advices for airline decision-making.In this paper,the civil aviation passenger service data is preprocessed,and the feature that is useful is extracted from large-scale and high-dimensional data.Then our study include two part of jobs.Firstly,we propose a feature extraction model of civil aviation suspicious order based on L1 regularized logistic regression.This model use L1 sparsity to extract the feature,and the regression coefficient can study the related relationship between the features.Secondly,we study the civil Aviation order data horizontally,and build a kmeans of ELM feature space to clustering the civil aviation passengers.By using ELM to map the sample feature nonlinearity to the high-dimensional elm feature space,and then using Non-Negative Matrix Factorization to reduce the feature dimension.ELM feature mapping and Non-Negative Matrix Factorization can improve data partitioning ability of kmeans.we finally divide the civil aviation passenger order data into five classification.After the data mining,we proposal a suspicious order recognition of civil aviation based on PSO-ELM feature mapping KNN classification.This algorithm uses the ELM feature mapping to map the feature to the ELM feature space,then uses the PSO to select the best mapping weight and hidden layer bias,then use the KNN to classify.The experiment result show that the classification ability is better than other traditional classification algorithms on suspicious order recognition.