Research On The Evaluation Method Of Airline Customer Satisfaction Based On ISSA-XGBoost

Improving customer satisfaction is an important measure for the development of airlines,and adopting appropriate customer satisfaction evaluation methods is crucial.The XGBoost algorithm is an excellent ensemble algorithm in machine learning that can effectively complete tasks such as classification and regression.Therefore,this thesis constructs an airline customer satisfaction evaluation model based on XGBoost.Sparrow search algorithm(SSA)has shown good performance in global optimization problems,so this thesis selects SSA to optimize the XGBoost algorithm parameters to improve evaluation accuracy.This thesis improves on the shortcomings of SSA by using the improved sparrow search algorithm(ISSA)to optimize the parameters of XGBoost algorithm,and constructs a customer satisfaction evaluation model based on the improved sparrow search algorithm to optimize XGBoost algorithm parameters(ISSA-XGBoost).The model constructed in this thesis can classify the overall satisfaction of customers with airline services based on their relevant information,enabling airlines to take relevant measures for satisfied customers and neutral or dissatisfied customers.At the same time,airlines can also analyze the relevant factors that affect overall satisfaction and make improvement measures.The specific research work of this thesis is as follows:(1)Build a customer satisfaction evaluation model based on XGBoost.This thesis conducts data cleaning,analysis visualization,feature engineering,and other operations on a large amount of customer information data.Then,XGBoost algorithm and three other classification algorithms are used to construct a customer satisfaction evaluation model,achieving the classification of overall customer satisfaction.Subsequently,some evaluation indicators are used to evaluate the model.The results show that the model based on XGBoost algorithm outperforms the other three models in various indicators,with a final classification accuracy of 96.21% and an AUC value of 0.99.The model performs well.(2)Improved sparrow search algorithm.The sparrow algorithm is prone to falling into local optima in the later stage of optimization,so this thesis integrates the following strategies for improvement: introducing a Circle chaotic map to initialize the sparrow population;Introducing the sine cosine algorithm to update the discoverer’s position;Using an improved formula to update the position of the watchman.Subsequently,its performance was compared with that of the ordinary sparrow algorithm and other improved sparrow algorithms,and its applicability was verified through parameter optimization using ISSA on two datasets.The results show that the convergence accuracy and speed of ISSA are significantly better than those of SSA in 5 unimodal and 5 multimodal test functions,and the average and standard deviation of ISSA optimization results are also better than other improved algorithms.The ISSA proposed in this thesis has significantly better performance and certain applicability.(3)Build a customer satisfaction evaluation model based on ISSA-XGBoost.Due to the fact that traditional parameter optimization methods may not necessarily find the global optimal solution,this thesis uses ISSA to optimize the maximum number of trees,maximum depth of trees,and learning rate of the XGBoost algorithm.At the same time,other three optimization algorithms are selected for comparison.The results show that compared with other optimization algorithms,ISSA has the smallest optimal fitness.The classification accuracy of the model based on ISSA-XGBoost has reached 96.52%,which is 0.31% higher than before.Other evaluation indicators,such as accuracy and recall,have also improved to varying degrees.The AUC value of the model approaches 1,and the model performs well.