Analysis of Association Rules for Travelers Staying at ESG-Certified Hotels in Taiwan

Abstract

This study investigates the behavioral determinants of tourists’ selection of ESG-certified hotels by applying association rule mining to 895,962 valid records derived from datasets of the Ministry of Transportation and the Ministry of Environment. Tourists were classified into three groups based on length of stay. The results reveal strong associations between ESG hotel choice and factors such as gender, age, port of entry, transport mode, and arrival city, with prominent patterns including “Kaohsiung Port,” “Age 30–39,” and “Airplane.” This study offers both theoretical contributions and actionable policy implications, advocating data-driven strategies to advance sustainable hotel management and effectively engage high-potential market segments.

1. Introduction

In recent years, the rise of global sustainable development issues and the increasing awareness of environmental protection have led to the evolution of Corporate Social Responsibility (CSR) into more concrete and measurable ESG (Environmental, Social, Governance) management standards. Within the scope of Environmental, Social, and Governance (ESG) criteria, safety constitutes a fundamental pillar underpinning economic development and the well-being of communities. For instance, Huang et al. [1] employed Geographic Information Systems (GISs) to analyze trends in firefighting and rescue operations, incorporating variables such as evolving economic conditions and regional traffic flow. From the standpoint of corporate sustainable performance, these factors are subject to dynamic adjustment and expansion over time. Regarding social responsibility in conjunction with consumer rights and corporate governance, Lin et al. [2] proposed an optimal sales model for an online travel database digital navigation system. Their study investigated how information providers can secure equitable profits from both travelers and businesses. By integrating sustainability considerations, they addressed supply and demand challenges within the online travel database from the perspective of mechanism design, further examining the sensitivity of travelers’ optimal total reading time per unit time to various model parameters. The promotion of ESG has not only become an important criterion for measuring the sustainable performance of enterprises but has also gradually become a significant basis for influencing consumer choices and investor assessments of corporate value. The tourism industry, which heavily relies on natural resources, social participation, and service quality, is particularly sensitive to ESG issues, making it one of the most important practical fields in various countries’ sustainable policies.

This study focuses on travelers staying in ESG-certified eco-friendly hotels in Taiwan. The preferences and habits of travelers regarding accommodation choices in Taiwan require considerable attention and planning from hotel operators and regulatory authorities. Therefore, ESG-certified eco-friendly hotels can meet the needs of visitors to Taiwan who prioritize accommodation quality and align with global trends related to environmental issues. In Taiwan, the government actively promotes systems such as “Green Hotels” and “Eco-Friendly Certification Hotels,” attempting to assist hotel operators in enhancing energy-saving and carbon-reduction practices, improving social responsibility performance, and introducing governance mechanisms to create a sustainable accommodation system that meets international trends. The Ministry of Transportation and Communications and the Ministry of the Environment collaborate to integrate resources through platforms like “Accommodation Network” and “Net Zero Green Living Information Platform,” providing guidelines for hotel operators to apply for eco-friendly certification and enhancing information disclosure and transparency of green certifications.

However, as ESG sustainable hotel-related systems become increasingly refined, whether travelers consider this an important factor in their accommodation choices still requires further verification. Research by Hsu et al. [3] on travelers from Singapore, Malaysia, New Zealand, Australia, and other regions visiting Taiwan indicates that international travelers are more concerned about accommodation quality and service professionalism compared to price factors. This suggests that if ESG-oriented hotel products and marketing strategies are properly planned, it will help enhance overall travel quality and tourism competitiveness.

Previous research on ESG hotels has mainly focused on aspects such as financial performance, brand strategy, and employee management. For example, Chung et al. [4] used DEA data envelopment analysis to verify the positive impact of ESG implementation on operational efficiency, while Yu et al. [5] found that hotels with clear ESG attributes can enhance brand loyalty and customer satisfaction. Bae [6] further proposed a specific ESG evaluation framework for the hotel industry, including three major areas: environment, society, and governance, with a total of 20 indicators and 41 sub-indicators, addressing the limitations of traditional corporate ESG models that do not fit the practicalities of the service industry.

Moreover, current ESG assessments often focus on the corporate perspective, with less attention paid to consumer behavior—especially that of international travelers—and neglect the implicit relational logic in their process of choosing ESG hotels. From an industry perspective, an increasing number of international hotel brands, such as Marriott, InterContinental, Starwood, and Shangri-La, have begun to publish ESG reports, emphasizing their practices in environmental sustainability, community engagement, and corporate governance. However, whether there is an effective connection from the customer perspective remains an area for further research.

Therefore, this study aims to integrate ESG management theory with consumer behavior analysis and introduce data mining methods to systematically explore the behavioral patterns of travelers staying in ESG-certified eco-friendly hotels in Taiwan. Notably, past research on travelers’ choices of ESG hotels has often employed qualitative methods, such as surveys or interviews, making it difficult to grasp potential rules under large samples and multiple variables. Thus, this study utilizes association rule mining algorithms, combining the Apriori and IARM models for big data analysis, which helps extract potential rules from vast amounts of data and provides concrete recommendations for decision-makers.

The specific research objectives of this study are as follows:

(1)

To explore the potential correlations between the characteristics of travelers from different nationalities, genders, ages, and travel purposes and their choice of ESG-certified eco-friendly hotels.

(2)

To apply the Apriori and IARM association rule algorithms to mine potential rules between traveler characteristics and the selection of ESG-certified eco-friendly hotels.

(3)

To construct a data mining model of traveler behavior and ESG hotel selection, providing references for government, the accommodation industry, and tourism policy planning practices.

In terms of research contributions, this paper has three innovative aspects. First, it combines traveler data from the Ministry of Transportation and Communications with data from eco-friendly certified hotels, integrating empirical data on tourism behavior and sustainable accommodation; second, it introduces the IARM algorithm to strengthen the limitations of traditional association rule calculations in effectively ranking important rules; third, it attempts to reverse-engineer the selection preferences for hotel ESG certifications from the user’s perspective, supplementing the current ESG data structure primarily based on operator declarations and enhancing the consumer-side data foundation, providing a scientific basis for future sustainable tourism promotion.

This study aims to bridge the gap between sustainable development policies and smart tourism practices, using big data and machine learning methods to address the current shortcomings in consumer behavior aspects of ESG hotel research, thereby promoting the deep implementation of ESG concepts in the tourism industry.

2. Literature Review and Research Hypotheses

2.1. ESG Perspectives and Eco-Friendly Certification for Hotels

The development of the ESG system is a response to the shortcomings of traditional corporate social responsibility (CSR), emphasizing the overall responsibility of businesses towards the environment, society, and governance during their operations. With the rise of global sustainability awareness, international attention to corporate sustainability performance has shifted from a moral level to a strategic level. Particularly, the tourism and hospitality industry, being resource-intensive and highly service-oriented, has become a key practice area for the implementation of ESG. The hotel industry not only faces environmental issues, such as energy, water resource, and waste management, but also involves challenges related to employee rights, community participation, and corporate governance.

According to research by Bae [6], South Korea has developed a specific ESG assessment index for the hotel industry. This index was constructed based on a Delphi survey method, inviting 12 experts with backgrounds in the hotel industry to participate in two rounds of opinion collection, ultimately resulting in three major dimensions, 20 indicators, and 41 specific assessment items. The environmental dimension includes aspects such as pollution emission improvement rates, green transportation, and information disclosure; the social dimension covers employee protection, emotional labor management, and local community participation; and the governance dimension assesses board composition, ethical management, and decision-making transparency. This framework significantly addresses the limitations of traditional corporate ESG models that fail to reflect the characteristics of the service industry, providing tools for hotels to quantitatively assess their sustainability performance.

Abazaj [7] used the Mainport Hotel in the Netherlands as a case study to empirically explore the integration and challenges of ESG principles in practice. The study analyzed the impact of ESG system implementation on operational models through employee surveys and in-depth interviews with managers. The results indicate that raising employee awareness and enhancing governance system transparency are key to successfully promoting ESG; at the same time, actions such as energy and water conservation, food waste recycling, and sustainable menu planning also significantly reduce operational costs, creating positive cyclical benefits.

In Taiwan, the Ministry of Transportation and Communications and the Ministry of the Environment jointly promote the “Eco-Friendly Hotel Certification” system as an important policy tool for advancing ESG in the domestic accommodation industry. This certification is divided into three levels: “Gold,” “Silver,” and “Bronze,” covering multiple evaluation indicators, such as energy saving, carbon reduction, resource recycling, and social participation. This certification system has a three-tier differentiation mechanism, with clear evaluation standards that include 3.1 mandatory compliance items and 3.2 optional compliance items, fully corresponding to the three major aspects of ESG. The 3.1 items are basic thresholds, including air quality management, environmental policy formulation, establishment of energy and water resource baselines, employee environmental education, hazardous substance management, and waste classification systems; 3.2 includes bonus items such as the usage rate of energy-saving equipment, the proportion of water-saving toilets, the quantity of green procurement, community participation, and green action promotion. For example, restaurants not using endangered species as ingredients, hotels installing heat recovery equipment, or guest rooms using water-saving devices are all assessment criteria. Additionally, those who obtain CNS 14001 certification can be exempted from certain inspection items, encouraging hotels to integrate environmental management systems; the certification system also stipulates that hotel operators must continuously provide proof of environmental measures and encourages silver and bronze level operators to gradually move towards gold level standards, forming a positive cycle of self-improvement [8]. Through institutionalized management, the Environmental Protection Administration can establish consistent evaluation criteria, providing operators with clear bases for sustainable actions and offering consumers reliable choice information. According to data from the Ministry of the Environment, by the end of 2023, over 200 hotels in Taiwan had passed the certification, showing a stable growth trend year by year. Notably, the eco-friendly certification has become an important filtering criterion for some OTA platforms, such as Agoda and Booking.com, indicating a certain demand for this information from consumers.

However, from the perspective of classification systems, the current hotel star rating evaluations still focus heavily on hardware facilities and service processes, with insufficient emphasis on sustainability. Wszendybył-Skulska and Panasiuk [9] pointed out that the internal hotel classification standards in the European Union still primarily focus on traditional facility orientation, with only HSU (Hotel Stars Union) incorporating limited environmental indicators, most of which are non-mandatory conditions. In contrast, the local hotel classification system in Poland lacks clear environmental management requirements. Both have failed to adequately reflect the core spirit of sustainable development.

Kim et al. [10] also emphasized that the transparency and consistency of ESG information disclosure in the hotel industry still need to be strengthened. Their analysis of ESG reports from the world’s top 50 hotel groups found that while most companies are willing to disclose information, there is a lack of unified formats and standards, making it difficult for consumers and investors to make effective comparisons. They suggested that international organizations or governments should promote the establishment of ESG disclosure guidelines for the accommodation industry to enhance information quality and credibility.

In addition to the design of institutional and evaluation mechanisms, the internal execution strategies of hotels are also crucial. Ding and Tseng [11] found that if hotels in China can effectively integrate ESG strategies, they can achieve significant results in financial performance and customer satisfaction. The study utilized data from 348 hotel responses to verify the complementary relationship between the three dimensions of environment, society, and governance, which can jointly promote the enhancement of overall competitive advantage.

From the literature reviewed, it is evident that the development and implementation of the ESG (Environmental, Social, and Governance) system in the hotel industry have gradually matured. However, it still faces challenges, such as insufficient classification standards, divergent assessment mechanisms, and limited consumer awareness. Therefore, it is necessary to enhance the transparency of ESG information, establish unified assessment criteria for the hospitality industry, and utilize data science methods to explore potential preferences and choice logic regarding sustainability indicators from user behavior.

2.2. The Relationship Between Tourists and ESG Eco-Label Hotels Based on Association Rule Mining

In the field of tourism behavior and decision-making, understanding the patterns of tourists’ accommodation choices is crucial for formulating effective tourism strategies. Especially as ESG has become an important framework for assessing the sustainable operation and service quality of hotels, exploring the behavioral relationship between tourists and ESG eco-label hotels holds both practical and theoretical value. With the development of big data and smart tourism, data mining techniques, particularly Association Rule Mining (ARM), are increasingly being applied to empirical analyses of tourism behavior.

Versichele et al. [12] conducted pattern mining of tourist attractions in Belgium using Bluetooth tracking data and applied ARM to identify the preferences of different tourists in their travel routes, demonstrating that specific tourist behavior rules can be identified even in non-participatory data collection. The “visitation pattern map” established in the study provides a visual tool for exploring the flow of tourists in space, highlighting the application potential of association rules in tourism management. Similarly, Arreeras et al. [13] conducted Wi-Fi tracking in Hokkaido’s tourist areas and combined ARM to analyze tourists’ movement and selection tendencies within large natural attractions, effectively revealing tourists’ potential preferences under environmentally friendly and sustainable tourism orientations.

Applying ARM to more detailed consumer behavior analysis, Li et al. [14] further combined positive and negative association rules to analyze decision-making for outbound tourism from Hong Kong, showing that the symbolic and directional nature of association rules can enhance the depth of tourism data interpretation. Li et al. [15] utilized 72,890 review data from the TripAdvisor platform to analyze tourist preferences and co-occurrence attributes for 218 tourist attractions in China, successfully identifying three types of tourists: landscape-sensitive, crowd-sensitive, and price-sensitive. By combining association rules and clustering techniques, the study confirmed that ARM can be used for market segmentation and product strategy formulation for travelers.

Additionally, Pyo [16] used ARM to integrate market segmentation, positioning, and target markets, finding that ARM can not only construct the logical structure of tourist preferences but also reflect the correspondence between destination driving factors and marketing motivations. By enhancing demand forecasting capabilities through data mining, it provides concrete bases for destination management and marketing strategies.

Beyond tourism behavior, there are also studies that combine ARM with ESG perspectives to explore sustainable tourism management. For example, Tosporn et al. [13] used Wi-Fi scanning technology to track tourists’ movement trajectories in natural attractions and conducted point combination preference analysis through ARM, proposing suitable route designs and facility configurations for sustainable tourism. Jiang [17] employed the FP-Growth algorithm to explore the characteristics of rural tourism demand and differences in experience levels, extending the application of ARM to tourism landscape planning and demand level modeling.

In recent years, some studies have attempted to incorporate ESG participation behavior into measurements of tourist behavior. For instance, Shin et al. [18] conducted a mixed-method study on tourists from South Korea and the United States, developing a tourist ESG participation scale that covers dimensions such as environmental behavior, social behavior towards residents, and governance participation. The study confirmed that tourists’ ESG behaviors not only affect their travel experiences and destination loyalty but can also serve as indicators for estimating their choice of ESG eco-label hotels. This finding provides a variable foundation for the association rule mining model for selecting ESG hotels, addressing the current blind spots in ESG assessments centered on operators.

From an analytical technique perspective, the Apriori algorithm and its improved version, IARM (Important Association Rule Mining), have been widely used for mining various consumer behaviors. Weng Cheng-Hsiung [19] pointed out that the IARM algorithm can enhance the traditional framework of support and confidence through importance indicators (Imp.), further optimizing the ranking and selection of association rules, thereby improving the interpretability and practicality of mining results.

This study continues the aforementioned methodological context, using tourist attributes (such as nationality, gender, age, purpose of visit to Taiwan, mode of transportation, length of stay, etc.) as the antecedent (LHS) of the association rules and whether to stay in ESG eco-label hotels as the consequent (RHS), employing Apriori and IARM algorithms for rule mining and importance ranking. Through Lift and Imp. indicators, it aims to identify and compare positive and negative associations, with the expectation of establishing a behavior pattern model of practical value. Overall, applying association rule mining to the study of ESG eco-label hotels not only fills the existing research perspectives centered on operators or systems but also expands the application of tourism big data in consumer behavior and sustainable tourism. Furthermore, by mining the association rules between tourists visiting Taiwan and ESG hotels, it can assist governments and operators in designing more attractive sustainable products and policy mechanisms for specific groups, thereby achieving the dual goals of ESG and tourism development.

3. Research Framework and Methods

3.1. Design of Association Rule Algorithm

Association rules can be mined using the Apriori algorithm. The analysis of association rules primarily explains how rules calculated from support and confidence are filtered to remove all meaningless association rules, followed by the deletion of duplicate rules. Additionally, the correlation indicator (lift) is calculated to identify both positively and negatively correlated rules. This is followed by a discussion on how to use the correlation indicator (lift) to filter out relevant association rules and how to select important indicators (Imp.) after extraction.

The main purpose of this study is to mine association rules related to travelers staying in ESG-certified eco-friendly hotels in Taiwan. It explores the characteristics and association rules of travelers staying in these hotels, utilizing the Apriori algorithm for association analysis. The steps for mining association rules are divided into two main steps: first, identifying high-frequency itemsets, and second, generating association rules [15,19]. This study calculates support, confidence, and correlation based on the Apriori algorithm proposed by Agrawal and Srikant and further develops the Important Association Rule Mining (IARM) algorithm to calculate the importance of rules.

Although the importance indicator enhances the existing framework of support and confidence, it allows for the measurement of the significance of association rules based on the level of the importance indicator, thereby reducing the number of association rules. This helps to identify meaningful and relevant rules from the analysis of association rules related to travelers staying in ESG-certified eco-friendly hotels, highlighting important association rules. In other words, the support and confidence framework used by the Apriori algorithm can filter out most duplicate and meaningless rules, while the correlation indicator is applied to select relevant association rules, and the importance indicator (Imp.) is calculated for these relevant association rules to identify significant associations.

The steps of the IARM algorithm are developed based on the Apriori algorithm, and the designs of both algorithms are closely related: (1) using a repetitive method to count the support of high-frequency itemsets one by one; (2) using the identified highly correlated high-frequency itemsets to generate Positive Association Rules (PARs), with the validity of the rules measured against a confidence threshold; (3) generating Negative Association Rules (NARs) for subsequent filtering of important association rules. The IARM algorithm uses correlation coefficients to identify positive association rules and calculates the support and confidence of negative association rules; (4) generating important association rules: comparing the confidence of positive and negative association rules to determine the importance of the association rules, thereby filtering out important association rules. The IARM algorithm applies the concept of correlation coefficient (lift) to measure the correlation between the left-hand side (LHS) and right-hand side (RHS) of association rules, thereby enhancing the support and confidence framework of the Apriori algorithm [19]. In other words, the IARM algorithm uses the importance indicator (Imp.) to measure the significance of association rules, thereby strengthening the support and confidence-based framework of the Apriori algorithm.

3.2. Research Database

The database sources include the “Ministry of Transportation and Communications’ Database of Travelers to Taiwan for 2022,” the “Tourism Bureau’s Accommodation Network,” and the “Environmental Protection Administration’s Net Zero Green Living Database.” The completed inbound tourist data is used for mining, categorized based on data attributes into three main categories: hotel data, traveler entry data, and data on stays at eco-friendly hotels.

The analysis was conducted based on the nationality of travelers to Taiwan, the number of nights they stay, and various characteristics of the cities they enter. The hotel data, traveler data, and characteristics of stays at eco-friendly hotels are used for the Apriori calculations, with the primary focus on inbound tourists staying at eco-friendly hotels.

Table 1. Analysis of incoming passenger data to Taiwan for 2022 by category and number of visitors.

Item	Group	Subtotal	Code
Nationality of Incoming Passengers to Taiwan	Asia	638,464	Area-a
	America	105,244	Area-b
	Europe	68,558	Area-c
	Oceania	14,994	Area-d
	Africa	3423	Area-e
	Overseas Chinese	65,279	Area-
Age	Under 19	90,027	Age 20–24
	20–29	234,398	Age 25–29
	30–39	239,780	Age 30–34
	40–49	156,331	Age 35–39
	50 and above	175,426	Age 40 up
Gender	Female	371,940	Female
	Male	524,022	Male
Purpose of Visit for Incoming Passengers in 2022	Business	96,620	Times-01
	Tourism	254,686	Times-02
	Family Visit	85,921	Times-03
	Conference	5893	Times-04
	Study	14,269	Times-05
	Exhibition	1666	Times-06
	Medical	2015	Times-07
	Other	434,892	Times-08
Transportation Used by Incoming Passengers to Taiwan in 2022	Airplane	875,907	Air-01
	Ship	20,055	Sea-02
Length of Stay for Incoming Passengers in 2022	Group 1—1 to 4 nights	167,134	Length of Stay-a
	Group 2—5 to 90 nights	285,259	Length of Stay-b
	Group 3—More than 90 nights	309,580	Length of Stay-c
	No overnight stay	133,989	Non
Total		895,962

presents the analysis categories and statistics of the inbound traveler database for 2022, including nationality of travelers to Taiwan, age, gender, purpose of visit, mode of transportation to Taiwan, and number of nights stayed.

Table 2. Analysis of categories and number of visitors staying at eco-label hotels by tourists visiting Taiwan in 2022.

Item	Group	Subtotal	Code
Whether travelers in 2022 stay at hotels with environmental certification in Taiwan	Staying at an eco-friendly certified hotel	273,917	ESG-01
	Not staying at an eco-friendly certified hotel	622,045	ESG-02
Total		895,962
Number of nights stayed by travelers in eco-label hotels in Taiwan for the year 2022	Group 1—1 to 4 nights	96,526	Length of Stay—ESG1
	Group 2—5 to 90 nights	94,634	Length of Stay—ESG2
	Group 3—More than 90 nights	82,757	Length of Stay—ESG3
Total		273,917
Travelers staying in eco-friendly certified hotels in Taiwan for the year 2022	Northern Group	129,248	Area_ ESG 1
	Central Group	29,586	Area_ ESG 2
	Southern Group	75,266	Area_ ESG 3
	Eastern and Outlying Islands	39,817	Area_ ESG 4
Total		273,917
Port of Entry for Travelers Arriving in Taiwan in 2022	Keelung Port	610	1_O
	Taichung Port	5127	2_N
	Kaohsiung Port	10,559	3_P
	Other Ports	3759	4_Q
Total		20,055
Cities of Arrival for Travelers Entering Taiwan in 2022	Taipei	747,817	City_1
	Taoyuan	201,744	City_2
	Taichung	1936	City_3
	Kaohsiung	24,275	City_4
	Others	135	City_5
Total		875,907

provides statistics on the categories and number of tourists staying at eco-friendly hotels in Taiwan for 2022, including stays at eco-friendly hotels, number of nights stayed at eco-friendly hotels, regions of eco-friendly hotels, entry ports for travelers to Taiwan, and cities of entry.

This study collected relevant data on travelers to Taiwan staying at eco-friendly hotels for 2022 from the “Ministry of Transportation and Communications’ Database of Travelers to Taiwan,” the “Tourism Bureau’s Accommodation Network,” and the “Environmental Protection Administration’s Net Zero Green Living Database.” The research scope covers the areas of eco-friendly hotels in all 22 counties and cities in Taiwan, also utilizing statistical compilation data from the National Development Council of the Executive Yuan [20] for regional clustering. The 22 counties and cities in Taiwan are divided into four regional groups: Northern, Central, Southern, Eastern, and Outlying Islands. The Northern group includes Taipei City, New Taipei City, Keelung City, Hsinchu City, Taoyuan City, Hsinchu County, and Yilan County; the Central group includes Taichung City, Miaoli County, Changhua County, Nantou County, and Yunlin County; the Southern group includes Kaohsiung City, Chiayi City, Tainan City, Chiayi County, and Pingtung County; and the Eastern and Outlying Islands group includes Taitung County, Hualien County, Kinmen County, and Lienchiang County.

3.3. Research Procedure

This study designs two different types of data volume for data mining calculations, based on the size of the data. The first type is the nationality record data of inbound travelers from the “Inbound Traveler Database,” with a total of 895,962 hotel stays for analysis. The second type involves selecting the number of stays of tourists in eco-label hotels from the “Tourist Database of Eco-Label Hotels” to cluster the length of stay of travelers in Taiwan. Therefore, this study clusters the length of stay of travelers in Taiwan, dividing the number of stays in eco-label hotels into three categories: 1 to 4 nights (Length of Stay—ESG1, totaling 96,526 stays), 5 to 90 nights (Length of Stay—ESG2, totaling 94,634 stays), and over 90 nights (Length of Stay—ESG3, totaling 82,757 stays). The following explains the process:

Data Type 1: The total number of entries in the “Inbound Traveler Database” is 895,962.

Step 1: The database and itemset $I=\left\{I_1\right.,\dots ,\left.I_n\right\}$ are defined, where the itemset represents every possible item that may appear in the database. A database is composed of a group of inbound traveler data $T=\left\{T_1\right.,\dots , \left.T_m\right\}$, each being a subset of the itemset.

Step 2: The Apriori algorithm is used to calculate association rules. In this study, X represents the basic information of inbound travelers, their entry data, and the hotel selection data of tourists, while Y indicates the presence of stays in eco-label hotels (divided into two phases: Phase One—stayed, Phase Two—did not stay). This research aims to explore the relationship and strength between the presence of stays in eco-label hotels and the background data of tourists and eco-label hotel data. Therefore, the identified association rules must exceed both the minimum support of 0.1 and the minimum confidence of 0.1, with a minimum length rule constraint of 2. In this phase, a total of 9 association rules were found.

Step 3: Since the Apriori algorithm can repeatedly generate large itemsets and quasi-large itemsets, a pruning step is necessary to eliminate redundant rules. Thus, the 9 association rules were pruned to form quasi-large itemsets to reduce duplicate rule combinations. After removing redundant rules, there are 7 positive correlation rules (lift(X,Y) > 1), as shown in

Table 3. Six positive correlation rules from the second data mining of the database on tourists’ choice to stay in eco-label hotels, with importance indicators greater than 80%.

No.	LHS	=>	RHS	Supp.	Conf.	lift	count	Imp.
(1)	{Port of Entry = Kaohsiung Port}	=>	{Group1—1 night to 4 nights}	71.23%	84.26%	1.2521	1254	91.67%
(2)	{Age = 30–39}	=>	{Group1—1 night to 4 nights}	70.25%	88.81%	1.3006	1613	90.30%
(3)	{Staying at eco-friendly certified hotel}	=>	{Group1—1 night to 4 nights}	68.16%	86.77%	1.2963	1349	88.22%
(4)	{Mode of Transportation = Airplane}	=>	{Group1—1 night to 4 nights}	67.68%	87.98%	1.1425	1546	87.23%
(5)	{Gender = Female}}	=>	{Group1—1 night to 4 nights}	67.98%	86.08%	1.2345	1426	86.90%
(6)	{Entry City = Taipei}	=>	{Group1—1 night to 4 nights}	65.34%	95.69%	1.0170	2598	84.48%
(7)	{Northern Group}	=>	{Group1—1 night to 4 nights}	62.19%	75.69%	1.0085	1477	79.02%

Note: The Supp. value must be at least 0.6 (inclusive), and the Imp. value must exceed 80%.

.

Step 4: After listing the positive correlation rules, the negative correlation rules are calculated. The positive correlation rules {X⇒Y} and their corresponding negative correlation rules {$\overline{X}$⇒Y} are extracted, and their importance indicators Imp. (X⇒Y) are calculated. The importance indicators of the positive correlation rules are shown in Table 3.

$$importance\left(X\Rightarrow Y\right)=\mathit{log}{\displaystyle \frac{Confidence\left(X\Rightarrow Y\right)}{Confidence\left(\overline{X}\Rightarrow Y\right)}}$$

Data Type 2: The “Tourist Database of Guests Staying at Eco-Certified Hotels” categorizes tourists who choose to stay at eco-certified hotels based on the number of nights they stay in Taiwan.

Step 1: From the “Tourist Database of Guests Staying at Eco-Certified Hotels,” the total number of guests who chose to stay at the hotels is 471. This database consists of a collection of items (Itemset) $I=\left\{I_1\right.,\dots , \left.I_n\right\}$ based on the grouping of tourists according to their length of stay at eco-certified hotels. The grouping is divided into three categories: 1 to 4 nights (Length of Stay—ESG1, totaling 96,526 visitors), 5 to 90 nights (Length of Stay—ESG2, totaling 94,634 visitors), and over 90 nights (Length of Stay—ESG3, totaling 82,757 visitors). The database for the three groups of tourists based on their length of stay in Taiwan, $D_{Length of Stay - ESG1}, D_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$, represents the possible items that can appear in the database for each group. The overall database is composed of three databases, $D_{Length of Stay - ESG1}, D_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$, categorized by the length of stay attributes at eco-certified hotels, with the data $T=\left\{T_1\right.,\dots , \left.T_m\right\}$ of tourists choosing to stay at these hotels forming subsets ${\mathrm{T}}_{\mathrm{i}}$ of itemsets for each group.

Step 2: Calculate the number of nights stayed data, $D_{Length of Stay - ESG1}{, D}_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$, within the database that contains three attribute groups, setting the minimum support to 0.1 and the minimum confidence to 0.1. Taking 1 night to 4 nights as an example, ${\mathrm{S}}_{\mathrm{N}}$ support is defined as support ($X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1})$P($X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$) = |$X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$|/m, which indicates the proportion of the set $X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$ appearing in the 1 night to 4 nights database $D_{Length of Stay - ESG1}$; confidence is defined as ($X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$) = P($Y_{Length of Stay - ESG1}$│$X_{Length of Stay - ESG1}$) = |$X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$|/|$X_{Length of Stay - ESG1}$|, which indicates the probability of the set $X_{Length of Stay - ESG1}$ appearing given the presence of the set $X_{Length of Stay - ESG1}{Y}_{Length of Stay - ESG1}$, representing the strength of the entire rule. The discovered association rules must simultaneously exceed the minimum support of 0.1 (minimum support) and the minimum confidence of 0.1 (minimum confidence), with a minimum length rule constraint of 2. In this phase, for $D_{Length of Stay - ESG1}$, a total of 10 association rules were found; for $D_{Length of Stay - ESG2}$, a total of 7 association rules were found; and for $D_{Length of Stay - ESG3}$, a total of 9 association rules were found, as shown in

Table 4. Association rules $D_{Length of Stay - ESG1}, D_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$ for tourists choosing to stay in eco-certified hotels based on length of stay in Taiwan.

School Attributes_Cluster	Attributes of Stay Duration for Eco-Label Hotels	Rules Total	Delete Duplicate Rules	Available Number of Rules
$D_{SLength of Stay-ESG1}$	1 night–4 nights	10	5	5
$D_{SLength of Stay-ESG2}$	5 night–90 nights	7	4	3
$D_{SLength of Stay-ESG3}$	More than 90 nights	9	4	5

.

Step 3: Since the Apriori algorithm can repeatedly generate large itemsets and candidate itemsets, it is necessary to perform a pruning step to eliminate unnecessary rules. Therefore, the $D_{Length of Stay - ESG1}{, D}_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$-generated association rules will prune the large itemsets into candidate itemsets to reduce redundant and unnecessary rule combinations. As a result, the confidence level of 0.1 will be simplified to 0.2. After deleting the redundant rules, the following describes the $D_{Length of Stay - ESG1}, D_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$-generated positive association rules:

For $D_{Length of Stay - ESG1}$, a total of 10 rules were generated, and $D_N$ had 5 positive association rules (lift(X,Y) > 1), as shown in

Table 5. A total of 5 positive correlation rules (lift(X,Y) > 1) for $D_{SLength of Stay - ESG1}$ 1 night–4 nights.

No.	LHS	=>	RHS	Supp.	Conf.	lift	count	Imp.
(1)	{Entry Port = Kaohsiung Port}	=>	{Group1—1 night to 4 nights}	89.66%	91.87%	1.1473	1241	89.37%
(2)	{Mode of Transportation = Airplane}	=>	{Group1—1 night to 4 nights}	80.34%	91.86%	1.2025	1211	86.55%
(3)	{Check into an eco-friendly certified hotel}	=>	{Group1—1 night to 4 nights}	72.63%	89.53%	1.4608	1630	82.43%
(4)	{Age = 30–39 years old}	=>	{Group1—1 night to 4 nights}	69.57%	88.38%	1.5360	1261	85.03%
(5)	{Gender = Female}	=>	{Group1—1 night to 4 nights}	46.67%	85.07%	1.2930	1314	77.36%

Note: The Supp. value must be at least 0.6 (inclusive), and the Imp. value must exceed 80%.

. Next, we will discuss the association rules with a support level of 0.6 (60%) or higher, all of which have met the criteria. Therefore, there are a total of 4 positive association rules (lift(X,Y) > 1) with a support level of 0.6 (inclusive) or higher.

For $D_{Length of Stay - ESG2}$, a total of 7 rules were generated, and $D_M$ had 3 positive association rules (lift(X,Y) > 1), as shown in

Table 6. 3 Positive correlation rules (lift(X,Y) > 1) for $D_{SLength of Stay - ESG2}$ 5 nights–90 nights.

No.	LHS	=>	RHS	Supp.	Conf.	lift	count	Imp.
(1)	{Stay at an eco-friendly certified hotel}	=>	{Group2—5 night to 90 nights}	79.35%	89.68%	3.2417	1247	84.26%
(2)	{Port of entry = Kaohsiung Port}	=>	{Group2—5 night to 90 nights}	78.14%	90.30%	3.8429	1360	82.77%
(3)	{Gender = Female}	=>	{Group2—5 night to 90 nights}	71.30%	88.61%	4.3054	1374	81.66%

Note: The Supp. value must be at least 0.6 (inclusive), and the Imp. value must exceed 80%.

. Next, we will discuss the association rules with a support level of 0.6 (60%) or higher, all of which have met the criteria. Therefore, there are a total of 3 positive association rules (lift(X,Y) > 1) with a support level of 0.6 (inclusive) or higher.

For $D_{Length of Stay - ESG3}$, a total of 9 rules were generated, and $D_S$ had 5 positive association rules (lift(X,Y) > 1), as shown in

Table 7. A total of 5 positive correlation rules (lift(X,Y) > 1) for $D_{SLength of Stay - ESG3}$ over 90 nights.

No.	LHS	=>	RHS	Supp.	Conf.	lift	count	Imp.
(1)	{Gender = Female}	=>	{Group3—Over 90 nights}	82.97%	90.54%	1.6271	1834	83.47%
(2)	{Mode of Transportation = Airplane}	=>	{Group3—Over 90 nights}	82.03%	89.38%	1.4634	1687	82.76%
(3)	{Age = 30–39 years old}	=>	{Group3—Over 90 nights}	77.93%	85.75%	1.8630	1968	81.37%
(4)	{Port of Entry = Kaohsiung Port}	=>	{Group3—Over 90 nights}	69.83%	82.73%	1.6713	1683	80.61%
(5)	{City of Entry = Taipei}	=>	{Group3—Over 90 nights}	69.07%	80.34%	1.5769	1955	80.06%

Note: The Supp. value must be at least 0.6 (inclusive), and the Imp. value must exceed 80%.

. Next, we will discuss the association rules with a support level of 0.6 (60%) or higher, all of which have met the criteria. Therefore, there are a total of 5 positive association rules (lift(X,Y) > 1) with a support level of 0.6 (inclusive) or higher.

4. Research Results

4.1. Results

4.1.1. Data Type 1: “Database of Tourists Choosing to Stay in Eco-Certified Hotels” Shows a Total of 895,962 Stays in Eco-Certified Hotels

This study utilizes this data for data mining association analysis, establishing the relationship rules between the characteristics of tourists visiting Taiwan and their choice to stay in ESG eco-certified hotels based on the collected hotel data, tourist data, and data on stays in eco-certified hotels. In the first round of data mining, with a minimum conditional support (Supp.) of 0.1, confidence (Conf.) of 0.1, and a minimum rule length limit of 2, a total of 30 association rules were calculated. Among these, only the rules for stays of 1 to 4 nights (the number of nights tourists stay in Taiwan) showed significant results. However, due to the excessive number of rules, it was difficult to identify meaningful ones, prompting a second round of data mining.

In the second round of data mining, with a minimum conditional support (Supp.) of 0.1, confidence (Conf.) of 0.4, and a minimum rule length limit of 2, redundant rules were removed, resulting in a total of 2 association rules. Among the positively correlated association rules, those with a lift value greater than 1 and an importance index greater than 80% totaled 6 association rules. The results of the positive correlation rules are shown in Table 3, and the importance index greater than 80% is ranked from highest to lowest as follows: entry port = Kaohsiung Port, age = 30–39 years, stay in eco-certified hotels, mode of transportation = airplane, gender = female, entry city = Taipei.

4.1.2. Data Type 2: Grouping Tourists Choosing to Stay in Eco-Certified Hotels Based on Length of Stay in Taiwan

The number of tourists choosing to stay in eco-certified hotels for $D_{Length of Stay - ESG1}$ 1 to 4 nights is 96,526, for $D_{Length of Stay - ESG2M}$ for 5 to 90 nights is 94,634, and for over 90 nights is 82,757, totaling 895,962 instances. The attributes $X_N$ of the research database $D_{Length of Stay - ESG1}$ include tourist background information, hotel stay information, and information on choosing eco-certified hotels, with a focus on stays of 1 to 4 nights, comprising 4 attributes. In this study, the three attribute groups regarding the length of stay in eco-certified hotels, $D_{Length of Stay - ESG1}, D_{Length of Stay - ESG2}, D_{Length of Stay - ESG3}$, were analyzed with a minimum support threshold of 0.1, a minimum confidence threshold of 0.4, and a minimum rule length of 2. Consequently, for $D_{Length of Stay - ESG1}$, a total of 10 association rules were identified in this phase, $D_{Length of Stay - ESG2}$ had 7 association rules found in another phase, and $D_{Length of Stay - ESG3}$ had 9 association rules identified in yet another phase, as shown in Table 4.

Generate 10 rules for the first attribute $D_{Length of Stay - ESG1}$ (1 night–4 nights). There are a total of five positive correlation rules (lift(X,Y) > 1), as shown in Table 5. Next, we will discuss association rules with a support of 0.6 (60%) or higher. Therefore, there are four positive correlation rules (lift(X,Y) > 1) with a support of 0.6 (inclusive) or higher, listed in order of support from highest to lowest: “Port of entry = Kaohsiung Port, Mode of transportation = Airplane, Staying in eco-label hotels, Age = 20–24 years.” There are four items with an importance index greater than 80%, listed in order from highest to lowest: Port of entry = Kaohsiung Port, Mode of transportation = Airplane, Age = 30–39 years, Staying in eco-label hotels.

Generate seven rules for $D_{SLength of Stay - ESG2}$ (5 night–90 nights), with three positive correlation rules (lift(X,Y) > 1) as shown in Table 6. Next, we will discuss the association rules with a support of 0.6 (60%) or higher. Therefore, there are a total of three positive correlation rules (lift(X,Y) > 1) with a support of 0.6 (inclusive) or higher.

The three positive correlation rules (lift(X,Y) > 1) with a support of 0.6 (inclusive) or higher are listed in order of support from highest to lowest: “Staying at environmentally certified hotels, Port of entry = Kaohsiung Port, Gender = Female.” There are three items with an importance indicator greater than 80%, listed in order from highest: Staying at environmentally certified hotels, Port of entry = Kaohsiung Port, Gender = Female.

Generate nine rules for $D_{SLength of Stay - ESG3}$ (Over 90 nights), with five of them being positive correlation rules (lift(X,Y) > 1), as shown in Table 7. The association rules are discussed based on a support of 0.6 (60%) or higher. Therefore, there are a total of five positive correlation rules (lift(X,Y) > 1) with a support of 0.6 (inclusive) or higher.

Since there are five positive correlation rules (lift(X,Y) > 1) with a support of 0.6 (inclusive) or higher, the characteristics of the rules are listed in descending order of support as follows: “Gender = Female, Mode of Transport = Airplane, Age = 30–39 years, Port of Entry = Kaohsiung Port, City of Entry = Taipei.” Among these, there are five items with an importance index greater than 80%, listed in order from highest to lowest as Gender = Female, Mode of Transport = Airplane, Age = 30–39 years, Port of Entry = Kaohsiung Port, City of Entry = Taipei.

4.2. Discussion

The number of nights travelers stay in Taiwan is categorized into three attributes, and the data is further divided into overall and subgroups (1–4 nights, 5–90 nights, and over 90 nights). The data regarding tourists choosing to stay in environmentally certified hotels is discussed based on the following calculation results:

In terms of the overall data, to reduce redundancy and avoid unnecessary rule combinations that are hard to focus on, the more frequent confidence level (Conf.) of 0.1 is adjusted to focus on rules with a confidence level (Conf.) of 0.4 (40%) or higher, which yields the best focus effect. Therefore, the minimum support (Supp.) is set at 0.1, the confidence level (Conf.) at 0.4, and the minimum rule length is limited to 2. The positive correlation rules total six: “Port of entry = Kaohsiung Port, Age = 30–39 years, Stay in environmentally certified hotel, Mode of transport = airplane, Gender = female, City of entry = Taipei.” Further calculations of the importance index (Imp.) show that the higher the value, the more important the rule is, allowing us to determine which rules are relatively significant. The results indicate that the highest importance index is for the first group—1–4 nights—with Port of entry = Kaohsiung Port at 91.67%, followed by Age = 30–39 years at 90.30%, and the third highest is Stay in environmentally certified hotel at 88.22%. The lowest importance index is for City of entry = Taipei at 84.48%. Thus, for the overall data, the importance index (Imp.) is Port of entry = Kaohsiung Port, Age = 30–39 years, Stay in environmentally certified hotel, Mode of transport = airplane, Gender = female, City of entry = Taipei (first group—1–4 nights).

For subgroup data, for 1–4 nights, if the minimum support (Supp.) is 0.1 and the confidence level (Conf.) is 0.4, there are 10 related rules, of which 5 are positively correlated. If we consider association rules with a support of 0.6 (60%) or higher, there are four rules, listed in order of support from highest to lowest: “Port of entry = Kaohsiung Port, Mode of transport = airplane, Stay in environmentally certified hotel, Age = 30–39 years.” Continuing with the calculation of the importance index, there are four items with an importance index greater than 80%, listed from highest to lowest: Port of entry = Kaohsiung Port, Mode of transport = airplane, Age = 30–39 years, Stay in environmentally certified hotel.

For 5–90 nights, with a minimum support (Supp.) of 0.1 and a confidence level (Conf.) of 0.4, seven rules are generated, with three being positively correlated. Continuing with association rules with a support of 0.6 (60%) or higher, there are three rules. These are listed in order of support from highest to lowest: “Stay in environmentally certified hotel, Port of entry = Kaohsiung Port, Gender = female.” Continuing with the calculation of the importance index, there are three items with an importance index greater than 80%, listed from highest to lowest: Stay in environmentally certified hotel, Port of entry = Kaohsiung Port, Gender = female.

For over 90 nights, using the same rules, nine rules are generated. Continuing with association rules with a support of 0.6 (60%) or higher, there are five positively correlated rules, listed in order of support from highest to lowest: “Gender = female, Mode of transport = airplane, Age = 30–39 years, Port of entry = Kaohsiung Port, City of entry = Taipei.” Continuing with the calculation of the importance index, there are five items with an importance index greater than 80%, listed from highest to lowest: Gender = female, Mode of transport = airplane, Age = 30–39 years, Port of entry = Kaohsiung Port, City of entry = Taipei.

From the subgroup data, in terms of association rules and importance indices, Port of entry = Kaohsiung Port, Age = 30–39 years, Stay in environmentally certified hotel, Mode of transport = airplane, Gender = female, City of entry = Taipei are all important indicators, but there are slight differences in the details of these variables across the groups:

(1)

Port of entry: For 1–4 nights, 5–90 nights, and over 90 nights, all three groups have Port of entry = Kaohsiung Port as a related rule and important indicator.

(2)

Age: For 1–4 nights and over 90 nights, both groups have Age = 30–39 years as a related rule and important indicator.

(3)

Whether to stay in an environmentally certified hotel: For 1–4 nights and 5–90 nights, both groups have Stay in environmentally certified hotel as a related rule and important indicator.

(4)

Mode of transport used by travelers coming to Taiwan: For 1–4 nights and over 90 nights, both groups have airplane as a related rule and important indicator.

(5)

Gender: For 5–90 nights and over 90 nights, Gender = female is a related rule and important indicator; for 1–4 nights, Gender = female is a related rule.

(6)

City of entry: For over 90 nights, City of entry = Taipei is a related rule and important indicator.

By cross-referencing the overall data with the three subgroup data, it is found that Port of entry = Kaohsiung Port appears as a related rule and important indicator in both the overall and subgroup data. Kaohsiung Port, utilizing its extensive transportation infrastructure and accelerated airport clearance procedures, alongside government-endorsed sustainability certification frameworks and environmentally friendly port facilities, offers an optimal foundation for sustainable travel tailored to ESG-conscious travelers. Additionally, from the six association rules and important indicators in the overall data, it can be seen that Port of entry = Kaohsiung Port, Age = 30–39 years, Stay in environmentally certified hotel, Mode of transport = airplane, Gender = female, and City of entry = Taipei are all either related rules or important indicators in the subgroup data.

This research, employing a critical examination of association rule analysis concerning travelers’ preferences for eco-labeled hotels, identified methodological and data-related constraints as potential limiting factors. The findings indicate a tendency for female travelers aged 30 to 39 to select eco-labeled accommodations; however, this conclusion is subject to considerable limitations. Notably, pronounced cultural biases contribute to significant variations in environmental awareness across travelers from different national backgrounds. Specifically, Western travelers typically exhibit a stronger emphasis on sustainable tourism, whereas Asian travelers may prioritize cost considerations.

5. Conclusions and Recommendations

5.1. Research Conclusions

Drawing upon data from the Ministry of Transportation and Communications Database of Incoming Travelers to Taiwan, the Tourism Bureau Accommodation Network, and the Environmental Protection Administration’s Net Zero Green Living Database for the year 2022, this study identifies key correlation patterns and significant indicators related to national special education tourists who opt to stay in environmentally certified hotels. The principal findings are as follows:

(1)

Significant correlation rules and importance indicators include entry through Kaohsiung Port, age group 30–39, female gender, mode of transport by airplane, and entry city Taipei, all of which are strongly associated with tourists’ preference for ESG-certified accommodations. Analysis of the overall data and segmented clusters reveals that Kaohsiung Port consistently emerges as a critical factor, suggesting that most tourists entering Taiwan via this port demonstrate a clear intention to stay in environmentally sustainable hotels. Furthermore, tourists aged 30–39 who stay for 1–4 nights exhibit a pronounced engagement with ESG-certified hotels, reflecting not only goal-oriented behavior but also a meta-cognitive awareness regarding their accommodation choices.

Age 30–39 remains a salient indicator across both short-term (1–4 nights) and long-term (over 90 nights) stays, indicating that preferences for ESG-certified hotels vary across different age cohorts. This underscores the complexity of fostering ESG awareness and environmentally responsible behaviors among travelers, as such preferences evolve with age and awareness development. Additionally, the decision to stay in environmentally certified hotels itself serves as a significant correlation rule across various stay durations, highlighting tourists’ commitment to environmental preparation and sustainable practices aligned with ESG principles.

Gender, specifically female, is identified as a significant indicator in both overall data and segmented groups for stays ranging from 1 to over 90 nights, suggesting a predominance of female tourists in the ESG-certified hotel market. This finding prompts further inquiry into whether this trend reflects market demand or structural characteristics of incoming traveler demographics. The entry city Taipei also emerges as a significant factor for tourists staying over 90 nights, indicating that these travelers are more likely to engage actively in sustainable practices consistent with ESG values.

(2)

For tourists staying 1–4 nights, key association rules include entry port Kaohsiung, mode of transport by airplane, age 30–39, female gender, and staying in environmentally certified hotels. These factors exhibit high support and importance levels, underscoring their relevance in short-term stay decisions.

(3)

Among tourists staying 5–90 nights, the association rules and important indicators comprise staying at eco-label hotels, entry via Kaohsiung Port, and female gender. These variables demonstrate strong positive correlations with support levels exceeding 60%.

(4)

For stays exceeding 90 nights, significant association rules include female gender, mode of transport by airplane, age 30–39, entry port Kaohsiung, and entry city Taipei. Each of these factors exhibits high support and importance, indicating their critical role in long-term stay decisions at ESG-certified hotels.

5.2. Recommendations

(1)

Stakeholders and relevant enterprises should prioritize the development and implementation of policies that facilitate and promote stays at eco-label hotels. Providing diverse accommodation options that enhance tourists’ ESG awareness and practical environmental competencies is essential. Given the consistent significance of Kaohsiung Port as an entry point, policy planning should incorporate this variable to better address demand and optimize outcomes related to eco-label hotel stays.

(2)

Efforts should be made to encourage male travelers, particularly those staying between 5 and over 90 nights, to select eco-label hotels. This strategy aims to actively cultivate and improve environmental literacy among this demographic, addressing observed gender disparities in ESG-certified hotel patronage.

(3)

Initiatives should also target travelers outside the 30–39 age bracket who stay 1–4 nights, as well as male tourists not currently choosing eco-label hotels for short-term and long-term stays. Enhancing environmental literacy and promoting active selection of eco-label accommodations within these groups can foster broader commitment to sustainable tourism practices.

(4)

Future research should incorporate personal characteristics, needs, and segmentation of travelers choosing eco-label hotels into analytical models. The current study was limited by data constraints, which precluded deeper exploration of these factors. Subsequent investigations could benefit from supplementary surveys and expanded databases that capture tourists’ environmental perceptions, motivations, and demographic attributes. Comparative analyses of association rules across different environmental awareness groups would further elucidate the determinants influencing the choice of ESG-certified accommodations.

(5)

Practical Strategies for Promoting ESG Hotels: Recommendations for Enhancing ESG Label Visibility on OTA Platforms, Demographically Targeted Marketing, and AI-Driven Dynamic Pricing. Based on the findings from association rule analysis, this study proposes three targeted practical strategies to advance the promotion of eco-certified hotels. The key variables identified include Kaohsiung Port, female travelers aged 30 to 39, and specific lengths of stay, which serve as the foundation for developing precise marketing approaches.

• Enhancement of ESG Label Integration on Online Travel Agency (OTA) Platforms:

Prominent international OTA platforms, including Booking.com and Agoda, have integrated sustainable accommodation labels into their ranking algorithms, while Google’s search algorithm similarly prioritizes eco-friendly lodging options. It is advisable for Taiwanese accommodation providers to proactively pursue internationally recognized ESG certifications, such as GSTC, LEED, and ISO 14064. Given that government subsidies may cover up to 80% of certification expenses, operators are encouraged to prominently display their environmental credentials on these platforms to enhance search engine rankings and increase visibility.

• Tailored Marketing Strategies Based on Demographic Segmentation:

Empirical evidence indicates that female travelers aged 30 to 39 exhibit a stronger preference for hotels bearing environmental labels, and millennials demonstrate a markedly greater emphasis on sustainable travel practices. Consequently, it is recommended to develop targeted green travel packages specifically designed for this demographic cohort, complemented by social media marketing campaigns aimed at promoting sustainable tourism concepts across digital platforms. Furthermore, differentiated service offerings should be formulated in accordance with the duration of stay to address the distinct needs of various traveler segments.

• Implementation of AI-Driven Dynamic Pricing:

Drawing upon the case studies of Delta Air Lines and Marriott Group, it is advisable to adopt an AI-based dynamic pricing system. Such a system is capable of real-time analysis of fluctuations in demand, competitor pricing strategies, and environmental, social, and governance (ESG) preferences. This enables the formulation of personalized pricing models tailored to environmentally conscious consumers. Specifically, the system can optimize pricing for ESG-compliant room categories during periods of high demand, while promoting eco-friendly packages through targeted discounts in off-peak seasons.

In summary, this study’s data mining approach provides valuable insights into the profiles and behaviors of tourists selecting environmentally certified hotels in Taiwan. The findings offer a foundation for targeted policy development and future scholarly inquiry aimed at advancing sustainable tourism aligned with ESG principles.