Satisfaction Analysis of Airline Passenger Experience

Yiran Zhang

School of Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai, 200240, China

Keywords: Airline Satisfaction, XGBoost Model, SHAP Model, Visualization, Recommendations.

Abstract: With the rapid development of the aviation industry, airlines are increasingly focusing on passenger

satisfaction with their flight experience. Not only does this enhance brand loyalty, it may also enable travelers

to become effective evangelists for the brand. In order to deeply explore what affects air passenger satisfaction

and the key factors behind it, this article studied a data set from Kaggle containing more than 120,000 airline

passenger satisfaction scores. Through comparative analysis and verification, this paper selected the XGBoost

model and SHAP model from many models. These two models have shown significant effects and accuracy

in identifying and predicting factors related to overall satisfaction. At the same time, this article also uses a

variety of data visualization methods to show the differences in satisfaction among different types of

passengers and different routes. After detailed analysis of the visualization results and combined with market

conditions, this article puts forward a series of targeted improvement suggestions, aiming to help airlines

optimize services and improve satisfaction.

1 INTRODUCTION

With the development of the aviation industry and the

process of globalization, competition in the aviation

market has become increasingly fierce, and it is

particularly important for airlines to manage

passenger satisfaction. Aviation satisfaction will have

a strong positive impact on repurchase and

recommendation intentions (Mateja, 2017). By

measuring passenger satisfaction, you can analyze the

key factors that affect satisfaction, measure the

current passenger satisfaction level, and make

improvement strategies for products and services,

thereby increasing the number of loyal customers and

increasing market share and competitiveness. By

building a satisfaction prediction model, passengers

can be segmented into the market, different service

products can be launched for different passengers, big

data can be fully mined, and passenger satisfaction

can be improved.

With the development of computer technology,

many scholars use artificial intelligence methods such

as machine learning to predict customer satisfaction.

For example, through confirmatory factor analysis

(CFA) and structural equation modeling (SEM),

Chung found that airlines' corporate social

https://orcid.org/0009-0002-6809-9425

responsibility activities can significantly affect

passenger satisfaction, airline brand, and airline trust.

In addition, passenger satisfaction and airline trust

have a significant impact on airline loyalty (Chung,

2022). Spain's Mora and others have also revealed an

assessment of customer satisfaction and loyalty levels

of Colombia's low-cost and traditional airlines,

arguing that the low-cost model reduces passenger

satisfaction, thereby reducing loyalty (Garmendia，

2021). At the same time, many articles also use the

XGBoost model to analyze and predict satisfaction.

Xu et al. integrated XGBoost and SMOTE algorithms

to study which factors can be used to predict patient

satisfaction (Xu, 2022). Gaofeng Guan et al apply the

XGBoost model to rank the importance of each topic

variable according to satisfaction (Guan, 2022).

Whether in medicine, supply chain or service

industry, etc., it is not uncommon to use big data to

analyze customer satisfaction (Sao, 2022; Alsayat,

2023).

The research route of this article is divided into

four stages: data processing, model construction,

visual analysis, conclusions and suggestions. This

article intends to conduct research and analyze the

overall satisfaction scores of 120,000 airline

passengers, as well as the differences in the priorities

Zhang, Y.

Satisfaction Analysis of Airline Passenger Experience.

DOI: 10.5220/0012911400004508

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2024), pages 141-152

ISBN: 978-989-758-713-9

141

and personalized needs of different groups of

passengers for air travel. On this basis, conclusions

are drawn and recommendations are made.

2 DATA PROCESSING

First, the article will conduct a simple summary

analysis of the data. In order to facilitate data

processing, the author named the 14 services as

categories 1-14, as shown in table 1:

Table 1: Data column name.

Class Name

1 Departure and Arrival Time Convenience

2 Ease of Online Booking

3 Check-in Service

4 Online Boarding

5 Gate Location

6 On-board Service

7 Seat Comfort

8 Leg Room Service

9 Cleanliness

10 Food and Drink

11 In-flight Service

12 In-flight Wifi Service

13 In-flight Entertainment

14 Baggage Handling

Figure 1a: Average statistics.

Figure 1b:Variance Statistics.

The analysis found that in-flight service and

baggage handling had higher average ratings, while

the convenience of online reservations, gate location,

and in-flight Wi-Fi service had lower average ratings,

as shown in Figure 1a and Figure 1b. Wifi service had

lower average ratings. Meanwhile, departure and

arrival times have the largest variance for

convenience, while in-flight service and baggage

handling have the smallest variance.

At the same time, the overall satisfaction

distribution chart is shown in Figure 2:

The author found that the overall satisfaction level

is neutral or there are more dissatisfied passengers

than satisfied passengers, indicating that the overall

satisfaction level is not high and there are many

problems to be solved.

The correlation coefficient between different

types of services is small and the correlation is weak.

Therefore, different types of services can be regarded

as independent of each other, and only the impact of

each service on the overall satisfaction level is

considered, as shown in Figure 3.

3 MODELING ANALYSIS

3.1 Feature Correlation Analysis

In order to study the impact of each service on overall

satisfaction, the author calculated the correlation

coefficient between them and drew a heat map as

Figure 4:

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

142

Figure 2: Overall satisfaction distribution chart.

Figure 3: Heat map of correlation coefficients of various types of services.

Figure 4: Heat map of the correlation coefficient between each category of services and overall satisfaction.

Satisfaction Analysis of Airline Passenger Experience

143

It is easy to find from the chart that online check-

in, in-flight services, and comfort have higher

correlation coefficients and have a greater impact on

overall satisfaction. The convenience of online

reservations, boarding gate location, and departure

and arrival times have the lowest correlation

coefficients and have the smallest impact on overall

satisfaction.

3.2 Model Selection

In order to select the appropriate model, the author

uses five different models: Logistic Regression,

Random Forest, Adaptive Boost, Categorical Bayes,

and XGBoost to evaluate and compare the accuracy

of each service. The ratio of the number of test sets

and training sets is 1 :4, the Precision and Recall rates

of each model are in table 2.

The precision and recall rates of each model are

visualized as shown in the Figure 5.

Obviously, each value of the XGBoost model is

the best among the five models. Therefore, it is

judged that the XGBoost model meets the situation of

airline passenger satisfaction and is the most suitable

for processing the data set of this question.

3.3 XGBoost Importance Analysis

Using the XGBoost model to analyze the

importance of each service, the results are shown in

the Figure 5.

Table 2: Comparison of model performance indicators.

Model

Accuracy on

Training

Accuracy on

Test

ROC AUC

Score

Precision Recall

Logistic

ression

0.877097 0.877101 0.944868 0.860610 0.846040

2 Random Forest 0.929014 0.930794 0.977088 0.924671 0.910297

3 Adaptive Boost 0.907251 0.906715 0.963083 0.901406 0.876238

4 Categorical Bayes 0.893242 0.893093 0.949520 0.874839 0.871980

5 XGBoost 0.942293 0.951376 0.986617 0.943811 0.918020

Figure 5: Model performance index comparison chart.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

144

Figure 6: XGBoost model running results.

In Figure 6, it can be found that online check-in

and in-flight Wi-Fi service are the most important,

while food and beverages and cleanliness have the

least impact on overall satisfaction. Service

categories with high importance should be focused on

improving, while service categories with less impact

can be appropriately ignored.

3.4 SHAP Model Prediction

Figure 7: SHAP model running results.

Using the average SHAP value as an indicator of

feature importance, as shown in Figure 7. Hyunwoo

Park and Hyunju Jeon used the SHAP model to

identify characteristics of customer satisfaction

dimensions (Hyunwoo, 2022). It can be seen that the

in-flight Wifi service has the farthest radiation range

around zero point, indicating that it is the most

influential feature. Most of the points for items such

as food and beverages are piled at the zero point,

indicating that their importance is the lowest.

3.5 Summary

Through the construction and operation of XGBoost

and SHAP models, the author obtained the following

results: among various services, the importance of

online check-in, leg space service and in-flight Wifi

service are all at the forefront. Therefore, it is

recommended that airlines can reduce queuing time

and improve passenger satisfaction by providing a

more convenient online check-in process, including

advance seat selection, online check-in, etc.; provide

wider seat space to provide more comfortable To

improve the riding experience, you can also consider

adding adjustable seats and leg support facilities to

meet the needs of different passengers; provide stable

and high-speed in-flight wireless network

connections to meet the needs of passengers, and

Satisfaction Analysis of Airline Passenger Experience

145

regularly maintain and upgrade in-flight wireless

network equipment. to ensure proper functioning and

good coverage.

Through the above measures, higher priority

service categories can be improved, thereby

increasing overall satisfaction.

4 VISUAL ANALYSIS OF

DIFFERENT TYPES OF

PASSENGER

Sudhakar once distinguished between passenger

types and airline types to analyze customer

satisfaction (Sudhakar, 2020). This article will

analyze the satisfaction data of different types of

passengers in five aspects to determine which types

of passengers need to improve which service levels.

4.1 Analysis of Differences in

Satisfaction Among Passengers of

Different Ages

First, the data set of more than 120,000 passengers

was cleaned to remove data with a score of "0" in

various services. Then, a statistical chart is

constructed with the age of passengers as the

horizontal axis and the proportion of satisfaction

and neutral or dissatisfaction as the vertical axis, as

shown in the Figure 8.

In order to further analyze the differences in

satisfaction of passengers of different ages for

different categories of services, the average scores of

various services for each age were calculated and

visualized in three dimensions. At the same time, the

average scores of passengers for different categories

of services were calculated for different age groups

based on the difference in overall satisfaction. As

shown in the Figure 9 and Figure 10.

Figure 8: Proportion chart of overall satisfaction among travelers of different ages.

Figure 9: Scatter plot of average ratings of different services by travelers of different ages.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

146

Figure 10: Line chart of average ratings of different services by travelers of different age group.

Figure 11: Distribution chart of overall satisfaction of passengers by gender.

It can be found in the Figure that the satisfaction

and average ratings of people in the 40-60 age group

are higher than those of other age groups in each

rating item, while the satisfaction and average ratings

of people in the 0-20 age group are generally the

lowest. In addition, people over 60 years old have

lower ratings for in-flight services, in-flight

entertainment and WIFI services.

The reason for the lower ratings of passengers in

the 0-20 age group may be a lack of travel experience,

not accustomed to aviation services, or low tolerance,

and the rating standards are relatively stricter. At the

same time, for older passengers, there may be certain

difficulties in using new technological equipment and

services, and in-flight entertainment items may have

certain requirements for vision or hearing, and older

people may not be interested in the entertainment

content provided on board. For example, the movies

and TV programs provided by airlines may not meet

the preferences of passengers over 60 years old,

resulting in lower ratings.

It is recommended that airlines should understand

the needs and preferences of passengers of different

age groups and provide personalized services. For

example, provide more entertainment facilities and

Internet services for young passengers, and provide

more comfort facilities and auxiliary services for

elderly passengers; Providing more fast food and

snack options for younger travelers and healthier and

more digestible dining options for older travelers;

Provide entertainment facilities such as movies,

music and games for young passengers, and provide

reading materials and relaxation space for elderly

passengers. Airlines can use the above measures to

improve the satisfaction of passengers of different age

groups and provide a better travel experience.

4.2 Analysis of Differences in

Satisfaction among Passengers of

Different Genders

Data is extracted based on three elements: gender,

overall satisfaction, and average rating of each

service, as shown in the Figure 11 and Figure 12.

As can be seen from the Figure 11 and Figure 12,

the proportion of male and female passengers who

express satisfaction is almost equal to the total

number. The proportion of female passengers who

express neutrality or dissatisfaction is slightly higher

than that of males, but there is no significant

difference between the two. And in the average

ratings of various services, gender did not cause

Satisfaction Analysis of Airline Passenger Experience

147

significant differences in ratings. It can be determined

that gender has no significant impact on overall

satisfaction and ratings.

4.3 Analysis of Differences in

Satisfaction among Travelers of

Different Customer Types

As shown in the Figure 12 and Figure 13.

As you can see from Figure 13 and Figure 14, the

number of return travelers is approximately four

times the number of first-time travelers. Among first-

time travelers, the number of neutral or dissatisfied

travelers is about three times the number of satisfied

travelers, while among returning travelers, the

number of neutral or dissatisfied travelers is roughly

equal to the number of satisfied travelers. Among the

average ratings of various services, except for the

three categories of boarding gate location, baggage

handling and in-flight services, the average ratings of

returning passengers are slightly lower than those of

first-time passengers. The average ratings of other

returning passengers are significantly higher than

those of first-time passengers. Among them, the

average ratings of returning passengers are

significantly higher than those of first-time

passengers. The difference between space services

Figure 12: Histogram of average ratings of passengers of different genders.

Figure 13: Histogram of average ratings of travelers by different customer types.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

148

Figure 14: Distribution chart of overall satisfaction of passengers by different customer types.

Figure 15: Distribution chart of overall passenger satisfaction for different flight destinations.

and online check-in is most obvious.

The reason for the above differences may be that

passengers who fly with this airline again may have

more travel experience. They have clearer

understanding and expectations of the airline's

aviation services. Compared with first-time

passengers, they can be more familiar with the

relevant service procedures and better evaluate and

compare the service quality of different airlines.

Therefore, it is recommended that airlines should

focus on and solve the needs of passengers

experiencing their own flights for the first time and

provide more information. For example, during the

booking process, provide first-time users with clear

information about the legroom size and configuration

of each seat to help them make a better choice;

Provide detailed boarding instructions on the airline's

website or app, including times, locations and steps,

to help first-time users understand the entire boarding

process.

4.4 Analysis of Differences in

Satisfaction Among Passengers

with Different Flight Destinations

As shown in the Figure 15 and Figure 16.

The number of business travelers is about twice

that of individual travelers, and the number of

satisfied business travelers is significantly higher than

the number of neutral or dissatisfied business

travelers. Among individual travelers, the number

who were neutral or dissatisfied was about nine times

the number who were satisfied. Except for the

convenience of departure and arrival times and check-

in services, the average ratings of business travelers

on other types of services are significantly higher than

those of individual travelers.

The above differences may be due to the fact that

airlines generally invest more resources and efforts in

providing services to business passengers, including

more spacious and comfortable seats, additional

benefits (such as priority boarding, VIP lounge

access) and higher-level dining and entertainment

facilities, etc. After business travelers receive better

services, they pay more attention to items that waste

Satisfaction Analysis of Airline Passenger Experience

149

Figure 16: Histogram of average passenger ratings for different flight destinations.

Figure 17: Distribution chart of overall passenger satisfaction for different flight destinations.

their precious time, such as flight delays and

complicated check-in procedures.

Therefore, it is recommended that airlines

continue to provide high-end services to business

travelers while striving to ensure that flights take off

and land on time to reduce the inconvenience and

pressure on business travelers' schedules. And

improve business travelers’ evaluation of the

convenience of flight times through effective

operations and on-time flight execution; At the same

time, attention should also be paid to investing in

personal passenger services and cultivating passenger

loyalty to improve the situation where the number of

neutral or dissatisfied passengers far exceeds that of

satisfied passengers.

4.5 Analysis of Differences in

Satisfaction among Passengers of

Different Cabin Classes

As shown in the Figure 17 and Figure 18.

As can be seen from the above Figure 17 and 18,

the overall satisfaction of business class passengers is

much higher than that of economy class and premium

economy class passengers, while the overall

satisfaction of premium economy class passengers is

slightly higher than that of ordinary economy class

passengers. In contrast, business class passengers also

have higher demands for convenience in departure

and arrival times than economy class passengers.

Therefore, the average rating of the corresponding

items is lower. The average rating of the boarding

gate location and on-board Wi-Fi service is similar to

that of economy class passengers. Therefore, it is

recommended to improve the quality of in-flight

wireless networks, optimize boarding gate guidance

services or provide shuttle buses, and strengthen

services for economy class passengers to solve the

problem of high proportion of neutral or dissatisfied

passengers.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

150

Figure 18: Histogram of average passenger ratings for different flight destinations.

5 CONCLUSIONS

The article selects the XGBoost model, combine with

SHAP model prediction and analysis to find that

online boarding, legroom services and in-flight Wi-Fi

services are the most important to be focused on

improvement. At the same time, this article also puts

forward relevant improvement suggestions for these

services. Such as optimizing advance seat selection

and online check-in procedures, providing wider seat

space and more comfortable seating experience,

regular maintenance and upgrade of in-flight wireless

network facilities, etc.

This article analyzes the differences in satisfaction

among different types of passengers in five aspects.

In terms of age, satisfaction among younger and older

age groups is lower, mainly reflected in online check-

in, in-flight services and in-flight entertainment.

Therefore, it is recommended that airlines understand

the preferences of passengers of all ages and provide

personalized services. In terms of gender, there is no

significant difference in the overall satisfaction of

male and female passengers, and the ratings of

various services are similar. Therefore, it is

determined that gender has no significant impact on

passenger satisfaction. In terms of customer types, the

satisfaction of returning passengers is significantly

higher than that of first-time passengers, specifically

in terms of legroom service and online check-in.

Therefore, it is recommended that airlines improve

more detailed boarding guides and instructions on

facilities use to help first-time passengers better adapt

to the flight. Regarding the purpose of the flight, the

satisfaction of business travelers is much higher than

that of individual travelers, and business travelers

also rate most services higher than individual

travelers. It is recommended that airlines strive to

ensure the punctuality and convenience of flights and

enhance their investment in services for individual

travelers. In terms of cabin class, the satisfaction level

of business class passengers is obviously the highest,

followed by the satisfaction level of premium

economy class passengers, and the lowest satisfaction

level of economy class passengers. This is also

consistent with the actual situation. It is

recommended to focus on optimizing the in-flight

Wifi service with lower ratings.

Overall, passengers' overall satisfaction is low,

with the convenience of online reservations, gate

location and in-flight Wifi receiving the lowest

ratings among various services. It is recommended

that airlines optimize services based on actual

conditions, especially in categories with low ratings

and high importance, to improve passenger

satisfaction.

REFERENCES

Alsayat, A, 2023. Customer decision-making analysis

based on big social data using machine learning: a case

study of hotels in Mecca. Neural Comput Applic. 35,

4701–4722.

Chung, S., 2022. The influence of CSR on airline loyalty

through the mediations of passenger satisfaction, airline

brand, and airline trust: Korean market

focused. Sustainability, 14(8), 4548.

Guan, G., Liu, D., Zhai, J. 2022. Factors influencing

consumer satisfaction of fresh produce E-commerce in

the background of COVID-19—A hybrid approach

based on LDA-SEM-XGBoost. Sustainability, 14(24),

16392.

Satisfaction Analysis of Airline Passenger Experience

151

Garmendia Mora, J. C., Tavera Morales, D. A,

2021.Aerolíneas bajo costo y convencionales: la

satisfacción y lealtad de sus clientes. Revista

Universidad & Empresa, 23(41), 1-26.Mateja K, 2017.

An investigation of customer satisfaction with low-cost

and full-service airline companies, Journal of Business

Research,80, 188-196

Hyunwoo P, 2022. Hyunju J, The Dynamics of Customer

Satisfaction Dimension based on BERT, SHAP, and

Kano Model, IFAC-PapersOnLine, 55, 10, 2384-2389.

Sao, A., Sharma, N., Singh, S., Vishwas Yelikar, B.,

Bhardwaj, A. 2023. Examining Challenges In The

Adoption of Big Data In Health Care Institutions And

Its Impact On Patients Satisfaction: An empirical study

in Delhi, India: - Global Conference on Emerging

Technologies, Business, Sustainable Innovative

Business Practices, And Social Well-Being. Asia

Pacific Journal of Health Management, 18.

Sudhakar, S., Gunasekar, S, 2020. Examining online ratings

and customer satisfaction in airlines. Anatolia, 31, 2,

260–273.

Thekkoote, R. 2022. Understanding big data-driven supply

chain and performance measures for customer

satisfaction. Benchmarking, 29(8), 2359-2377.

Xu, Y., Wu, G., Chen, Y, 2022. Predicting patients'

satisfaction with doctors in online medical communities:

An approach based on XGBoost algorithm. Journal of

Organizational and End User Computing, 34(4), 1-17.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

152