Research On Characteristics Of Aviation Customer Segments Based On Data Mining Techniques

In order to realize the differentiated management of airline customers and improve the company’s market competitiveness,airlines need to analyze customer needs.Since customer business data contains a large amount of redundant data with excessive order of magnitude and complex dimensions,ordinary statistical analysis methods cannot effectively analyze such a complex data set,so it has important theoretical and practical significance to use data mining techniques to analyze customer demand and perform airline customer segmentation data.Currently most airlines are relatively rough for frequent flyer segmentation,but other traditional industries such as finance and telecommunication have some achievements in using data mining technology for customer segmentation and feature research,using data mining technology for effective customer segmentation and customer group feature analysis is an urgent development topic nowadays.This article aims to combine customer relationship management theory and data mining technology to establish a segmentation model for airline customers,reveal the characteristics and needs of different customer groups through analysis of airline customer data,and propose corresponding customer service strategies.This article improves the traditional RFM model,proposes the RFMDP model with civil aviation characteristics,and constructs a customer group index system.In the selection of clustering methods,this article uses K-means and K-means++ clustering algorithms to cluster customers,and integrates the four evaluation methods of Elbow Method,Silhouett Coefficient,Davies-Bouldin index and Calinski-Harabasz Index to The optimal number of clusters is found to solve the defect that the K-value(number of clusters)of the algorithm needs to be set artificially,and the result with the best clustering effect is selected as the final result of customer segmentation.Finally,combining the airline customer business data,the algorithm model was analyzed empirically,and it was found that the curve of Elbow Method was too smooth to draw a conclusion.the curve of Calinski-Harabasz Index was more volatile.However,the three evaluation indexes of Silhouette Coefficient,Davies-Bouldin index and Calinski-Harabasz Index can be found that the results of K-means++ clustering algorithm are better and more relevant,and also the best number of clusters is five customer groups,and according to the different customer groups Based on the characteristics of different customer groups,we propose corresponding differentiated service and marketing improvement suggestions.