*2.3. Determination of Kano Model by Berger*

There were five options for functional and dysfunctional questions, including "Like", "Must-be", "Neutral", "Live with", and "Dislike". According to the choices of the people investigated, the Kano type of this general demand was obtained, as shown in Table 5. Among them, the Kano types obtained according to the dysfunctional and functional questions correspond to the Kano type in Table 2. Then, the principle of relative majority was adopted to summarize all Kano types of each demand item and determine the final Kano type of each demand item. Berger [33] proposed an improved Kano category, which is defined as follows:

**Table 5.** Dysfunctional and functional questions [33,34].


<sup>1</sup> Note: Q is a suspicious result due to incorrectly set question or misunderstanding of respondent.

If f (O+A+M) > f (I+R+Q), the Kano type is the highest frequency type among O, A, and M;

If f (O+A+M) < f (I+R+Q), the Kano type is the highest frequency type among I, R, and Q;

If f(O+A+M) = f (I+R+Q), the Kano type is the highest frequency type among O, A, M, and I;

where f (X) is the frequency of the demand type.

#### *2.4. Determination of TS and CS of the Kano Model*

The second part is mixed class analysis, proposed by Newcomb [35], which includes two indicators (as shown in Equations (1) and (2)), TS (Total Strength) and CS (Category Strength). If the TS value ≥ 0.6 and the CS value ≤ 0.06, it indicates that the demand belongs to the mixed class H, and the two mixed types, namely the two types with the highest frequency, should be explained.

$$\text{TS} = \text{f}(\text{O} + \text{A} + \text{M})/\text{f}(\text{O} + \text{A} + \text{M} + \text{I} + \text{R} + \text{Q})\tag{1}$$

$$\text{CS} = \frac{\max\{\text{f}(\text{O}), \text{f}(\text{A}), \text{f}(\text{M}), \text{f}(\text{I}), \text{f}(\text{R}), \text{f}(\text{Q})\} - \text{second } \max\{\text{f}(\text{O}), \text{f}(\text{A}), \text{f}(\text{M}), \text{f}(\text{I}), \text{f}(\text{R}), \text{f}(\text{Q})\}}{\text{f}(\text{O} + \text{A} + \text{M} + \text{I} + \text{R} + \text{Q})} \tag{2}$$

#### *2.5. Determination of Better-Worse Coefficient of Kano Model*

The third part is the analysis of the Better–Worse coefficient, which is calculated by the percentage of each demand to the classification. The Better–Worse coefficient is used to indicate the influence degree of whether the demand is satisfied or not for the respondents. The better coefficient is an "increased satisfaction coefficient", indicating that when the demand/function is satisfied, the user's satisfaction will be improved. The closer the coefficient is to 1, the more obvious the improvement in satisfaction will be. The Worse coefficient is "dissatisfaction coefficient after elimination", indicating that when the demand/function is eliminated, the user's satisfaction will decrease. The closer the coefficient is to −1, the more obvious the decrease in satisfaction will be. The Better coefficient and Worse coefficient are calculated as follows (Equations (3) and (4)):

$$\text{Better coefficient} = \frac{\text{f}(\text{O}) + \text{f}(\text{A})}{\text{f}(\text{O} + \text{A} + \text{M} + \text{I})} \tag{3}$$

$$\text{Worse coefficient} \,\text{t} = -\frac{\text{f}(\text{O}) + \text{f}(\text{M})}{\text{f}(\text{O} + \text{A} + \text{M} + \text{I})} \tag{4}$$

The evaluation of the Average Satisfaction Coefficient (ASC) based on the Better–Worse coefficient was proposed by Park [36], which is strongly correlated with the Better and Worse coefficients and can reflect the priority degree in functional design. The higher the ASC value (as shown in Equation (5)), the higher priority the change demand/function should be.

$$\text{ASC} = \frac{|\text{Better}| + |\text{Worse}|}{2} \tag{5}$$

If we plot a four-quadrant graph based on the Better–Worse coefficients of the demand factors, the "Better" coefficient is along the *X*-axis and the absolute value of the "Worse" coefficient is along the *Y*-axis. The origin is the average value of the absolute value of the "Better" coefficient and the "Worse" coefficient at each point.

#### **3. Findings from the General Demand Analysis**

As shown in Figure 5, 78.51% of respondents indicated that they needed software/ systems to guide them to specific shops. As shown in Figure 6, 47.19% of respondents indicated that they were not satisfied with the current indoor navigation signage/system.

**Figure 6.** "Satisfaction with existing indoor navigation" fan chart.

#### *3.1. Reliability and Validity Analysis*

In this survey, 502 questionnaires were issued in total, of which 498 were effective, accounting for 99.20%. SPSS 22.0 software (IBM SPSS Statistics, Norman H. Nie & C. Hadlai (Tex) Hull & Dale H. Bent Chicago, IL, USA) was used for reliability and validity statistics. The gender, age, and occupation of the respondents are shown in Table 6 and Figures 7–9. The gender, age, and occupation of the respondents in the questionnaire conform to the actual access situation of the shopping mall, and also conform to the conclusion of Wu [37] on customers of different ages in user surveys of a shopping mall. Due to economic ability and desire, users are mainly concentrated between 18 and 49 years old. Professional office workers with stable economic incomes accounted for the largest proportion. Therefore, the questionnaire is in line with the actual situation and the general results of previous studies.

**Table 6.** Sex and age cross-comparison table.


**Figure 7.** Gender frequency.

**Figure 8.** Age frequency.

**Figure 9.** Occupation frequency.

**Finding 1:** *The user profile of the indoor navigation system of the shopping mall is 18–49-year-old non-retired users with a certain economic ability and frequent access to different shopping malls*.

As a basic index of a questionnaire survey, reliability and validity are divided into two levels. Reliability is used to test the internal stability and consistency of the evaluation results, while validity is used to judge whether the results obtained from the questionnaire can accurately represent the evaluated functional requirements. The reliability measurement indexes include Cronbach's Alpha value, the halved coefficient, retest reliability, etc. The validity measurement is factor analysis, and the KMO value and Bartlett sphericity

test value are used in this paper. As shown in Table 7, Cronbach's Alpha of the functional questions is between 0.7 and 0.8 with good reliability, the KMO value is between 0.7 and 0.8, and Bartlett's sphericity test corresponds to a *p* value < 0.05, indicating that the results of functional questions in this questionnaire are suitable for factor analysis. Cronbach's Alpha of dysfunctional questions was between 0.7 and 0.8, with good reliability, the KMO value was between 0.7 and 0.8, and Bartlett's sphericity test corresponded to a *p* value < 0.05, indicating that the results of dysfunctional questions in this questionnaire were suitable for factor analysis.

**Table 7.** Reliability and validity analysis of forward and reverse problems of Kano model.


#### *3.2. Correlation Analysis*

Correlation analysis can determine the degree of correlation between variables and judge the degree of correlation between basic personal information and objective data and subjective data in the questionnaire, using the Pearson correlation coefficient and significance SIG value. The correlation analysis framework of this questionnaire is shown in Figure 10.

**Figure 10.** Framework for correlation analysis.

#### 3.2.1. Correlation Analysis from Objective Data

As shown in Table 8, there is a certain correlation between respondents' basic information and objective data. There is a significant difference between "shopping mall frequency" and "age" at 0.05 (double tail), and an extremely significant difference between "occupation" at 0.01 (double tail). The correlation between "familiar with indoor map of shopping mall", "age", and "occupation" is significantly different at 0.05 (double tail). The correlation between "whether can find the shortest path", "gender", and "age" is significant at 0.01 (double tail), and the correlation between "occupations" is significant at 0.05 (double tail). In addition to the correlation between characteristics (personal basic information) and state variables (objective data) in Table 7, there is also a correlation between status variables (objective data). The r value of "frequency of shopping mall" and "familiarity with indoor map of shopping mall" is 0.377, with a SIG value of 0.000, and the correlation is very significant at 0.01 (double tail). The r value and SIG value of "are you familiar with the indoor map of the shopping mall" and "can you find the shortest path to the destination" are 0.297 and 0.000, showing a very significant difference in the correlation at 0.01 (double tail).


**Table 8.** Correlation analysis between characteristics (personal basic information) and status variables (objective data) 1.

<sup>1</sup> r is Pearson correlation coefficient. p is significance sig value. Blue indicates significant difference. Red indicates very significant difference.

Figure 11 shows the basic information of "gender" and the frequency of the personal data contrast, according to the Pearson correlation coefficient and the significant SIG value of "gender" and "if you can find the shortest path". Males' perceptions of finding the shortest path were evenly balanced, while more than half of females felt they were unsure whether they would find the shortest path.

**Figure 11.** *Cont*.

**Figure 11.** Frequency comparison between "Gender" and three objective data. (**a**) Shopping mall frequency; (**b**) Be familiar with the indoor map of shopping mall; (**c**) Whether the shortest path can be found.

Figure 12 shows the comparison between the frequency of "age" and personal data. According to the Pearson correlation coefficient and significance SIG value, it can be seen that "age" is correlated with personal data related to shopping behavior. Although the proportion of respondents aged 30–49 is 1.23 times higher than that of respondents aged 18–29, their frequency of going to the mall more than once a week is 1.97 times higher than that of respondents aged 18–29, indicating that respondents aged 30–49 visit shopping malls more frequently. The frequency of respondents aged 18 to 29 years old visiting shopping malls mainly ranges from once a week (8.83%) to once every two weeks (9.03%) to less than once a month (10.04%), which is related to different work and rest schedules for different age groups. Young users need to consider academic and economic pressures, while middleaged users have more disposable time and consumption demands. Secondly, in terms of familiarity with the indoor map of the shopping mall, all age groups are concentrated in "not very familiar", and the proportion of "very familiar" with the indoor map is less than 2%. This again highlights the need for a properly designed indoor navigation system. Similar to "whether familiar with store indoor map", the 18-to-29-year-old group of respondents' answers to "if you can find the shortest path" focus on "no" (15.66%) and "uncertainty" (14.26%). Respondents in the 30–49 age group, however, concentrated on "yes" (15.26%) and "not sure" (19.48%), which were related to lifestyle fixation and familiarity with frequent mall visits. Respondents generally reflected that they could not find the shortest route in a new shopping mall. In addition, 55.18% of the respondents aged over 50 answered "yes", indicating that the elderly tend to go to a fixed shopping mall. They often go to the nearest shopping mall to buy daily necessities and are quite familiar with the route.

**Finding 3:** *Users of different age groups go to shopping malls with different frequencies, which is closely related to the lifestyle, economic strength, and work schedule of different age groups*.

**Finding 4:** *Users that frequently go to particular malls think they are familiar with the indoor map of the shopping mall, but still need a navigation system when they go to a new mall*.

**Finding 5:** *Users over the age of 50 tend to go to particular shopping malls due to habits and physical limitations, etc., and they do not use new navigation systems. Therefore, they are not the target users of indoor navigation systems in shopping malls*.

**Figure 12.** Frequency comparison between "Age" and personal data related to shopping behavior. (**a**) Shopping mall frequency; (**b**) Be familiar with the indoor map of shopping mall; (**c**) Whether the shortest path can be found.

Figure 13 shows the frequency comparison between "occupation" and personal data related to shopping behavior. According to the Pearson correlation coefficient and significance SIG value, it can be seen that "occupation" is correlated with personal data related to shopping behavior. The frequency of students visiting shopping malls is concentrated at least once every two weeks, which is similar to the related results of "age". The frequency of visiting shopping malls of the self-employed and retired respondents was relatively balanced. Office workers accounted for the largest proportion of respondents, with a frequency of mainly once a week. Secondly, in terms of whether they are familiar with the indoor map of shopping malls, students and office workers answered "not very familiar"; in contrast, 26% of office workers chose "relatively familiar", which also matched the middle-aged survey respondents' "regular life" and "regular shopping malls" mentioned above. In the part concerning "whether the shortest path can be found", 64.23% of the students believe that they cannot find it or are not sure, while 75.28% of the office workers focus on "yes" and "not sure", which also matches the previous conclusion.

**Figure 13.** *Cont*.

**Figure 13.** Frequency comparison between "Occupation" and personal data related to shopping behavior. (**a**) Shopping mall frequency; (**b**) Be familiar with the indoor map of shopping mall; (**c**) Whether the shortest path can be found.

**Finding 6:** *Similar results of "occupation" and the personal data related to shopping behavior can be corroborated with "age"*.

### 3.2.2. Correlation Analysis from Subjective Data

As shown in Table 9, there is a certain correlation between people's basic information and data related to shopping behavior. The correlation between "need navigation software/system" and "occupation" is significant at 0.05 (two tails). In addition to the correlation between characteristics (personal basic information) and demand variables (subjective data) in Table 8, there is also a correlation between demand variables (subjective data). The r value of "whether you need navigation software/system" and "Yes or no you are satisfied with the current navigation system" is −0.156, the SIG value is 0.000, and the correlation is significant at 0.01 (double tail).

**Table 9.** Correlation analysis between characteristics (personal basic information) and demand variables (subjective data) 1.


<sup>1</sup> r is Pearson correlation coefficient. p is significance sig value. Blue is significant difference.

Figure 14 shows the frequency comparison of "gender", a person's basic information, and the personal data related to shopping behavior, while Figure 15 shows the frequency comparison of "age", a person's basic information, and personal data related to shopping behavior. According to the Pearson correlation coefficient and significance sig value, it can be seen that "gender" and "age" are not correlated with them.

**Figure 14.** Frequency comparison between "Gender" and personal data related to shopping behavior. (**a**) Do you need a software system to guide you to the destination store? (**b**) Are you satisfied with the current signage/system for indoor navigation?

#### **Finding 7:** *"Gender" and "Age" are not correlated with personal data related to shopping behavior*.

Figure 16 shows the basic information of "professional" respondents and the frequency of personal data related to shopping behavior contrast, according to the Pearson correlation coefficient and the significant sig value of "professional" and "whether need navigation software/system". Except for the retired group, the respondents of other occupational types are not satisfied with the current indoor navigation signage or system, especially students (10.84%) and office workers (22.69%).

**Finding 8:** *Satisfaction with indoor navigation systems is low among students and office workers, while retirement groups do not care. This also verifies that the retired group is not the target user of the indoor navigation system*.

**Figure 16.** Frequency comparison of "Occupation" and personal data related to shopping behavior. (**a**) Do you need a software system to guide you to the destination store? (**b**) Are you satisfied with the current signage/system for indoor navigation?

Students and office workers are selected as the main users. Young people, limited by their academic and economic abilities, tend to choose uncertain shopping malls, and they are more willing to try different shopping malls, having a greater demand for indoor navigation. Middle-aged people have a more fixed life trajectory and are very familiar with the indoor map of a specific shopping mall, but they also have a greater demand for indoor navigation when going to a shopping mall they have never been to.

**Finding 9:** *There is a certain correlation between the characteristics of users and their behavior in shopping malls*.

#### *3.3. Kano Model Classification*

As shown in Table 10, the Kano type of the five demands is consistent with the modified Kano types of Berger.

**Table 10.** Frequency statistics and classification of indoor navigation demand factors in shopping malls.


#### *3.4. Mixed Classification*

As shown in Table 11, combined with the Kano model classification and mixed classification, the five demands of this questionnaire are all non-mixed types, which are consistent with the Kano model classification. "Vertical elevator first" belongs to the Attractive quality (A), while "Escalator first" belongs to the Indifferent quality (I). Both vertical elevators and horizontal escalators are important means of vertical transportation in shopping malls. Users prefer to take the vertical elevator because they think the vertical elevator is faster

and more convenient while consuming less energy. Some users stated that they are afraid of the height of the horizontal escalator.

**Table 11.** Mixed analysis of indoor navigation demand factors in shopping malls.


**Finding 10:** *Although vertical elevators are more difficult to find than horizontal escalators, users prefer to use vertical elevators, which may be related to the user's belief that the vertical elevator is faster*.

"Passing by a particular type of store" belongs to the Attractive quality (A), while "Bypass specific types of shops" belongs to the Indifferent quality (I). This shows that users do not accept unnecessary detours, but only detours due to congestion.

**Finding 11:** *The users prefer that the navigation system passes the desired space when reaching a certain destination, but users who are not interested will tend to avoid certain spaces*.

#### *3.5. Better–Worse Coefficient Analysis*

Table 12 shows the Better–Worse coefficient and ASC values of indoor navigation demand factors in shopping malls. According to AEC values, the priority of the five demands can be obtained: "Avoid crowded/emergency sections" > "Vertical elevator first" > "passing by specific types of shops" > "Escalator first" > "Bypass specific types of shops".

**Table 12.** Analysis of Better–Worse coefficient and ASC value of indoor navigation demand factors in shopping malls.


The four-quadrant diagram is drawn according to the Better–Worse coefficient of the demand factors for indoor navigation in shopping malls, as shown in Figure 17. "Avoid crowded/emergency roads" is located in the first quadrant, which is an expected requirement and should be given first priority in functional design. "Passing by specific types of shops" and "Vertical elevator first" are in the second quadrant, which should also be given priority. "Bypass specific types of shops" and "Escalator first" are located in the third quadrant, which are undifferentiated requirements and can be ignored in functional design.

In terms of functional design, it can be concluded that users prefer vertical elevators for vertical transportation. In addition, it is generally recognized as necessary to avoid congested road sections or emergencies. In terms of going to a certain destination and passing the rest of the space, the users tend to think that it is necessary to pass a certain type of store that they want to pass through, but it is not necessary to make a detour to avoid passing through a certain type of store.

**Figure 17.** Better–Worse coefficient quadrant diagram.

**Finding 12:** *The three indoor navigation system functions identified in this paper are* "*Avoid crowded/emergency roads*", "*Vertical elevator first*", *and* "*Passing by a particular type of store*".

#### **4. Practical Implementations of Findings from the Survey**

Based on literature research and the questionnaire analysis of the Kano model, it can be concluded that "Avoid crowded/emergency roads" is an important demand, while "Vertical elevator first" and "Passing by specific types of shops" are expected general demands when designing the indoor navigation system of shopping malls. "Bypass specific types of shops" and "Escalator first" are undifferentiated requirements and may not be considered. Table 13 shows answers to "When planning the path of indoor navigation system, you want it to consider: \_\_\_\_\_\_" in the questionnaire. The survey results are similar to those of the Kano model. "Avoiding crowded/emergency sections", "Vertical elevator first", and "Passing by specific types of shops" account for the largest proportion, while "Bypassing specific types of shops" and "Escalator first" account for less than 40%.

**Table 13.** Percentage of indoor navigation considerations (multiple choice) in the questionnaire.


In addition to the analysis of the five general demands, the questionnaire also set the following question: "What are your requirements for indoor navigation in shopping malls/what do you think the most convenient and effective indoor navigation system should have: \_\_\_\_\_". A total of 141 valid answers were collected. The users' expectation of the system is being clear and easy to use, having detailed precision, and real-time accuracy. In terms of specific needs, some users mentioned the demand to identify shopping mall activities, indicate the location of toilets, and provide a variety of options and non-graphic designs by floor.

This paper determines demand functions of three levels, as shown in Table 14, including the function level, use level, and effect level. The function level is to realize three exciting needs and expectation needs and can choose whether to implement specific functions through a personalized interface. The interface should be simple and clear, with navigation made easy enough even for people with a weak sense of direction. At the effect level, new demands are put forward for the specific environment of shopping malls. For

example, the path congestion index needs to change in real time, and even predict future congestion based on past long-term data. It is necessary to solve the three-dimensional navigation results of the shopping mall from the two aspects of positioning and navigation, rather than a two-dimensional plane. More consideration should be given to special spaces such as toilets, vertical elevators, and horizontal escalators. Multiple paths can be provided if the user desires.


**Table 14.** Classification and design of shopping mall interior navigation requirements.

Based on the determined system functions, the shopping mall indoor navigation system as shown in Figure 18 was designed. The dotted box in Figure 16 is the preparatory work for the indoor navigation system of the shopping mall. First, based on the BIM model, the 3D road network is obtained through IFC and stored in the form of a matrix. Meanwhile, a database of indoor images of shopping malls is established, and 200 photos were selected for each of the four different types of scenes of elevators, atriums, gates, and ordinary indoor space. Subsequently, computer vision was used to determine whether a location in the mall is a congested route. Then, the multi-source heterogeneous information such as the congestion situation, shops on the intended route, three-dimensional coordinates, and vertical elevators are integrated into the shortest path routing algorithm. The parameters are continuously adjusted to obtain the most suitable route evaluation algorithm. Among them, the login page of the indoor navigation system of the shopping mall is shown in Figure 19, which provides the basic information of the user. In addition, the functional page design of the shopping mall indoor navigation system is shown in Figure 20. The user scans the QR code of the nearest store through the mobile phone, and their three-dimensional coordinates can be obtained in the background. On the mobile phone terminal, the user enters the destination, the shops they intend to pass through, whether to avoid crowded road sections, and whether vertical elevators are preferred, and then the optimal path calculated by the path planning algorithm can be obtained. The mobile terminal of the mobile phone presents the visual effect of the path, and displays the important store nodes of the path to improve readability and understandability.

**Figure 18.** Shopping mall indoor navigation system.


**Figure 19.** Shopping mall indoor navigation system login page.

**Figure 20.** Shopping mall indoor navigation system function page.

#### **5. Conclusions**

This paper provides a detailed analysis of users' general demand for indoor navigation systems for shopping malls. Three important functions that need to be considered in the design of shopping mall indoor navigation system are obtained. According to the questionnaire, this paper also drew some interesting conclusions, which will benefit further studies on indoor navigation systems for shopping malls: (1) Both vertical elevators and horizontal escalators are important means of vertical transportation in shopping malls. Although vertical elevators are more difficult to find than horizontal escalators, users prefer to use vertical elevators, which may be related to users' beliefs that the vertical elevator is faster. (2) Users prefer that the navigation system can indicate the passage to the desired space when reaching a certain destination, such as buying milk tea and other beverages along the way when going to the cinema, but users who are not as interested will avoid certain spaces. (3) This paper finds the correlation between "gender", "age", "occupation", and users' behaviors in shopping malls. For example, due to the singularity of moving lines, retired groups do not care about the indoor navigation signs of shopping malls. (4) Retirees over the age of 50 are not the target users of indoor navigation systems in shopping malls.

There are still some deficiencies in this paper. The functions formulated in the questionnaire survey are still insufficient, and the multi-source heterogeneous data fusion and application in the design of shopping mall indoor navigation systems are not fully explored. In future work, more in-depth research will be carried out on these two aspects.

**Author Contributions:** Y.X.: Methodology (lead), formal analysis, investigation, writing—original draft. H.D.: Resources, validation, writing—review and editing, funding acquisition. Y.D.: Conceptualization (lead), methodology, writing—review and editing, funding acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Guangdong Science Foundation (Grant No. 2022A1515010174), support from the State Key Lab of Subtropical Building Science, South China University of Technology (No. 2022ZB19), and the support from the Guangzhou Science and Technology Program (No. 202201010338).

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


**Jian Xu 1,2,3, Muchun Li 1,3,4,\*, Dandan Huang 1, Yuxin Wei <sup>1</sup> and Sijia Zhong <sup>1</sup>**


**Abstract:** The purpose of this study is to explore the occupants' subjective evaluation of the indoor environmental quality (IEQ) of hotels with the same physical environment and different decoration styles, and to reveal the influence of different decoration styles on the subjective evaluation of the indoor environmental quality. The study found a hotel with three mainstream styles of modern simple style, British pastoral style, and modern Japanese style, and adopted standard rooms with the same area, pattern, lighting, orientation, and decoration cost. The only variable controlled was the decoration style, and the subjective feelings of customers on the physical environment were investigated. Based on the literature and 604 online comments, the researchers designed a questionnaire and collected 710 effective questionnaires for empirical analysis. The analysis results of KH coder and SPSS software (Chicago, IL, USA) show that -1 the light environment in the indoor environment (including indoor natural lighting, lighting and other influencing factors) and non-light visual factors (including indoor color matching, plant layout, closeness to nature, decoration texture, space materials, decoration atmosphere and other factors) has the greatest impact on the subjective evaluation of decoration style, especially on the subjective evaluation of modern simple indoor environment. -2 Light environment, air quality and non-light visual factors play a key role in the subjective evaluation of the indoor environment of the British pastoral-style hotels. -3 The light environment, thermal environment and non-light visual factors are the most sensitive to the subjective evaluation of the indoor environment of modern Japanese-style hotels. -4 Thermal environment, light environment, acoustic environment, air quality environment and non-light visual factors have the greatest impact on the subjective evaluation of the hotel indoor environment. Based on the findings, this study puts forward some suggestions to improve the interior environment of the hotel with different decoration styles to improve the quality and attractiveness of the hotel.

**Keywords:** hotel indoor environment; decoration style; text analysis; questionnaire survey; subjective evaluation; comparative study; China
