Subjective Product Evaluation System Based on Kansei Engineering and Analytic Hierarchy Process

Product evaluation is very important for product improvement and development, and subjective product evaluation determines customer’s evaluation of products to some extent, so the purpose of this study is to establish a reasonable subjective product evaluation system. In this study, we comprehensively determine the evaluation indexes based on Kansei engineering (KE), establish an overall product evaluation system by using analytic hierarchy process (AHP), and establish the subjective product evaluation system by classifying the evaluation indexes in the overall product evaluation system into “objective evaluation index” and “subjective evaluation index”, removing the objective evaluation indexes, and retaining the subjective evaluation indexes. Additionally, we select some modern chairs as experimental samples to verify the reliability and validity of this subjective product evaluation system by means of questionnaires. The experimental results show that, in this subjective product evaluation system, the subjective evaluation of the product is positively correlated with the “favorite” level of the product in comprehensive evaluation, and negatively correlated with the “least favorite” level of the product in comprehensive evaluation, indicating that this subjective product evaluation system realizes a symmetry between subjective product evaluation and comprehensive product evaluation. Therefore, it can be concluded that this subjective product evaluation system based on KE and AHP proposed in this study has reliability and validity, and can be used for product evaluation to judge the popularity of products and enhance the competitiveness of products.


Introduction
The trend of product development has already changed towards the consumer-oriented; namely, the consumer's feeling and needs are recognized as invaluable in product development for manufacturers [1,2]. The "physical" performance of the product is no longer the only focus, and the "spiritual" performance of the product is getting more and more attention [3]. The products on the market must not only meet the basic physical and physiological needs of customers, but also satisfy the spiritual and emotional needs of customers [4][5][6].
Kansei engineering (KE), as a representative method of emotional design, has been widely used in the design field since its introduction [7]. Different from conventional emotional design methods, KE is a consumer-oriented ergonomics-based product design support technology that transforms vague emotional needs and images into design elements of products [3,8]. Up to now, there have been a lot of researches related to KE. Chang and Chen established a KE model that can integrate the interrelations of constituent elements to determine consumer emotional perception and preferences in a case regarding steering wheel design [9]; Cucuk et al. conducted research to determine the elements that must be considered in desk and chair design for elementary school, and KE is used in this research to capture select the perceptual phrases suitable for the research objects for analysis, which has a large workload and limits the extended application of the research results. In view of the drawbacks of the above studies, this study has made corresponding improvements: (1) based on the whole product field, this study explores various evaluation factors of the product under the four attributes of the product to establish the evaluation systems of this study, which makes the application scope of this study wider; (2) under the four attributes of products, this study summarizes the evaluation indexes of products, considering not only the concrete product elements such as product shape, color, and material, but also the abstract product elements, such as cultural connotation and social influence, which makes this study more comprehensive and reasonable; (3) this study comprehensively considers the factors influencing product evaluation based on KE, establishes an overall product evaluation system by using AHP, and then obtains the subjective product evaluation system by eliminating the objective evaluation indexes in the overall product evaluation system and retaining the subjective evaluation indexes, which effectively reduces the research errors and makes the research more reasonable; (4) this study simplifies the traditional KE, instead of searching for suitable perceptual phrases from a large number of perceptual words, it sets the perceptual phrases corresponding to all subjective evaluation indexes as "weak-strong", which makes the application scope of the research result wider.

Research Framework
This study consists of three parts, as shown in Figure 1. In part 1, KE and AHP are used to construct an overall product evaluation framework, that is, a hierarchy structure model. There are four levels in this framework, including the overall evaluation level, attribute level, aspect level, and evaluation index level. Then, the AHP is used to calculate the relative weight of each element at each level to obtain the overall product evaluation system. In part 2, the overall product evaluation system is divided into a subjective product evaluation system, and an objective product evaluation system. In part 3, 10 representative chair products are selected as experimental samples, and then appropriate subjects are selected to participate in the experiment to verify the validity and rationality of this subjective product evaluation system.

Kansei Engineering
KE is founded by Professor Mitsuo Nagamachi at Hiroshima University in the 1970s [1,37]. It aims at the implementation of the customer's feeling and demands into product function and design. Kansei is a Japanese word, which means the customer's psychological feeling, as well as embracing physiological issues. KE is defined as "translating the customer's kansei into the product design domain" [1,2,38]. In the field of industrial design, it expresses people's perceptual image of "things" quantitatively and semi-quantitatively, and associates it with product design characteristics, so as to realize the perceptual feeling of "people" (including consumers, designers, etc.) in product design and design products that meet the sensory expectations of "people" [3]. Usually, we use the semantic differentials (SD), developed by Osgood and his colleagues [39], as a main technique to grasp the consumer's kansei. SD is a basic research method, which reflects the user's perception on the Likert scale (usually the 5-point, 7-point, or 9-point scale) through the semantics of the research object (including the shape of the product, the color of the product, etc.), and then analyzes the laws by statistical methods [40]. The main purpose of this study is to establish a subjective product evaluation system that is as universal as possible, so our research object transitions from the whole of each product to the various influencing factors of each product (the influencing factors of each product are the same), and the evaluation of these influencing factors needs to be quantified uniformly. Therefore, on the basis of the theory of KE, we set the evaluation adjectives of these influencing factors as "weak-strong". The scale used in this study is the 5-point scale, as shown in Figure 2.

Kansei Engineering
KE is founded by Professor Mitsuo Nagamachi at Hiroshima University in the 1970s [1,37]. It aims at the implementation of the customer's feeling and demands into product function and design. Kansei is a Japanese word, which means the customer's psychological feeling, as well as embracing physiological issues. KE is defined as "translating the customer's kansei into the product design domain" [1,2,38]. In the field of industrial design, it expresses people's perceptual image of "things" quantitatively and semi-quantitatively, and associates it with product design characteristics, so as to realize the perceptual feeling of "people" (including consumers, designers, etc.) in product design and design products that meet the sensory expectations of "people" [3]. Usually, we use the semantic differentials (SD), developed by Osgood and his colleagues [39], as a main technique to grasp the consumer's kansei. SD is a basic research method, which reflects the user's perception on the Likert scale (usually the 5-point, 7-point, or 9-point scale) through the semantics of the research object (including the shape of the product, the color of the product, etc.), and then analyzes the laws by statistical methods [40]. The main purpose of this study is to establish a subjective product evaluation system that is as universal as possible, so our research object transitions from the whole of each product to the various influencing factors of each product (the influencing factors of each product are the same), and the evaluation of these influencing factors needs to be quantified uniformly. Therefore, on the basis of the theory of KE, we set the evaluation adjectives of these influencing factors as "weakstrong". The scale used in this study is the 5-point scale, as shown in Figure 2.

Analytic Hierarchy Process
AHP, developed by Thomas L. Saaty in 1971Saaty in -1975, is a MCDM method helping decisionmaker facing a complex problem with multiple conflicting and subjective criteria [16,20,42,43]. The core concept of AHP is decomposing a complex problem into a hierarchy structure (Figure 3), and assessing the relative importance of these criteria/indexes by pairwise comparison [44]. When setting up the AHP hierarchy with a large number of elements, the decision maker should attempt to arrange these elements in cluster so they do not differ in extreme ways [42,45,46]. In this study, we take the "evaluation indexes" in evaluation system as the indexes of the hierarchy structure, and the "evaluation indexes" are also the influencing factors in KE.
Result Z Level 1 Overall objective

Analytic Hierarchy Process
AHP, developed by Thomas L. Saaty in 1971Saaty in -1975, is a MCDM method helping decision-maker facing a complex problem with multiple conflicting and subjective criteria [16,20,42,43]. The core concept of AHP is decomposing a complex problem into a hierarchy structure (Figure 3), and assessing the relative importance of these criteria/indexes by pairwise comparison [44]. When setting up the AHP hierarchy with a large number of elements, the decision maker should attempt to arrange these elements in cluster so they do not differ in extreme ways [42,45,46]. In this study, we take the "evaluation indexes" in evaluation system as the indexes of the hierarchy structure, and the "evaluation indexes" are also the influencing factors in KE. core concept of AHP is decomposing a complex problem into a hierarchy structure (Figure 3), and assessing the relative importance of these criteria/indexes by pairwise comparison [44]. When setting up the AHP hierarchy with a large number of elements, the decision maker should attempt to arrange these elements in cluster so they do not differ in extreme ways [42,45,46]. In this study, we take the "evaluation indexes" in evaluation system as the indexes of the hierarchy structure, and the "evaluation indexes" are also the influencing factors in KE. Level   In AHP, after the hierarchy is constructed, we need to construct judgment matrices by pairwise comparing the importance of elements of the same level in the hierarchy based on the 1-9 fundamental scale (Table 1).

Intensity of Importance Definition 1
Equal importance 3 Weak importance of one over another 5 Essential or strong importance 7 Demonstrated importance 9 Absolute importance 2, 4,6,8 Intermediate values between the two adjacent judgments For example, A is an n • n pairwise comparison matrix (1), where a ij is the comparison between element i and j, and a ji = 1/a ij . In AHP, after the hierarchy is constructed, we need to construct judgment matrices by pairwise comparing the importance of elements of the same level in the hierarchy based on the 1-9 fundamental scale (Table 1).  [43].

Intensity of Importance Definition
Weak importance of one over another 5 Essential or strong importance 7 Demonstrated importance 9 Absolute importance 2, 4, 6,8 Intermediate values between the two adjacent judgments For example, A is an n · n pairwise comparison matrix (1), where a ij is the comparison between element i and j, and a ji = 1/a ij .
Then we need to calculate the maximum eigenvalue λ max of A.
In AHP, as priorities make sense only if derived from consistent or near consistent matrices, the consistency check must be applied. Only if the consistency ratio (CR) is less than 0.1, then the Symmetry 2020, 12, 1340 6 of 25 matrix can be considered as having an acceptable consistency, and reasonable, otherwise the matrix needs to be adjusted to acquire an acceptable consistency.
CI is the consistency index, RI is the random index, and the value of RI is shown in Table 2.

Overall Product Evaluation Framework
In the past, when people evaluated a product, they usually focused on three dimensions of "aesthetics", "function", and "commerce". However, with the satisfaction of people's material life, "human-orientation" has become a concern, and whether a product is designed from the perspective of "human" has become more and more important [47,48]. Arthur J. Pulos, a famous American design educator, also emphasized that the fourth dimension of "humanity" in design is more important than other dimensions. Therefore, in this study, we take "aesthetic attribute", "functional attribute", "commercial attribute", and "human-oriented attribute" as the base point, excavate the evaluation indexes that affect product evaluation under these four attributes based on KE, and then construct the overall product evaluation framework based on AHP. In the overall product evaluation framework, each attribute is divided into two to four aspects, and each aspect contains two to four evaluation indexes, as shown in Figure 4, where evaluation indexes with gray background are subjective evaluation indexes, and evaluation indexes with colorless background are objective evaluation indexes.

Weight Distribution of Each Level
In the overall product evaluation framework, the AHP is used to calculate the weight of each element relative to the element above it, and the weight of each element in the lowest level relative to the element in the highest level, to obtain the overall product evaluation system. Specific steps are as follows:

Construct judgment matrix
Several experts judge the relative importance of each element by pairwise comparison of elements of the same level in the overall product evaluation framework, and construct judgment matrices, according to Table 1.
For example, in this study, the comparison matrix of "Material Texture A 1 ", "Structure and Shape A 2 ", and "Overall Coordination A 3 " under "Visual Aspect X 1 " is shown in Table 3.

Weight Distribution of Each Level
In the overall product evaluation framework, the AHP is used to calculate the weight of each element relative to the element above it, and the weight of each element in the lowest level relative to the element in the highest level, to obtain the overall product evaluation system. Specific steps are as follows: 1. Construct judgment matrix Several experts judge the relative importance of each element by pairwise comparison of elements of the same level in the overall product evaluation framework, and construct judgment matrices, according to Table 1.
For example, in this study, the comparison matrix of "Material Texture A1", "Structure and Shape A2", and "Overall Coordination A3" under "Visual Aspect X1" is shown in Table 3.   That is, 2 Normalize A twice to obtain its eigenvector W, according to Equations (2) and (3) That is, W = 3. Calculate the maximum eigenvalue λ max of A, according to Equations (4) and (5) We can get  (6) and (7) and Table 2 Symmetry 2020, 12, 1340 , that is, this matrix has an acceptable consistency.

Summary
Under the Visual Aspect, the weights of "Material Texture", "Structure and Shape" and "Overall Coordination" are 0.1429, 0.1429, and 0.7142.
Following the above steps, we perform the remaining hierarchy single rankings and consistency checks, and obtain the following results.
The comparison matrix of "Aesthetic Experience A 4 " and "Cultural Connotation A 5 " under "Emotional Aspect X 2 " is shown in Table 4.

this matrix has an acceptable consistency. That is,
under the Emotional Aspect, the weights of "Aesthetic Experience" and "Cultural Connotation" are 0.7500 and 0.2500. The comparison matrix of "Material Prospective A 6 ", "Process Prospective A 7 ", "Material Advancement A 8 "and "Process Advancement A 9 " under "Technological Aspect X 3 " is shown in Table 5.
, λ max = 4, CI = 0, CR = 0 < 0.1, this matrix has an acceptable consistency. That is, under the Technological Aspect, the weights of "Material Prospective A 6 ", "Process Prospective A 7 ", "Material Advancement A 8 "and "Process Advancement A 9 " are 0.1250, 0.1250, 0.3750 and 0.3750. The comparison matrix of "Basic Function A 10 " and "Extended Function A 11 " under "Use Aspect X 4 " is shown in Table 6. under the Use Aspect, the weights of "Basic Function A 10 " and "Extended Function A 11 " are 0.8333 and 0.1667. The comparison matrix of "Adjustability A 12 " and "Convenience A 13 " under "Design Aspect X 5 " is shown in Table 7.
, this matrix has an acceptable consistency. That is, under the Design Aspect, the weights of "Adjustability A 12 " and "Convenience A 13 " are 0.1667 and 0.8333. The comparison matrix of "Durability A 14 " and "Stability A 15 " under "Quality Aspect X 6 " is shown in Table 8.

this matrix has an acceptable consistency. That is,
under the Quality Aspect, the weights of "Durability A 14 " and "Stability A 15 " are 0.1667 and 0.8333. The comparison matrix of "Mass Acceptance A 16 ", "Social Influence A 17 " and "Product Sales A 18 " under "Value Aspect X 7 " is shown in Table 9.
, λ max = 3.0660, CI = 0.0330, CR = 0.0569 < 0.1, this matrix has an acceptable consistency. That is, under the Value Aspect, the weights of "Mass Acceptance A 16 ", "Social Influence A 17 " and "Product Sales A 18 " are 0.1932, 0.7235 and 0.0833. The comparison matrix of "Material Cost A 19 ", "Process Cost A 20 " and "Transportation Cost A 21 " under "Price Aspect X 8 " is shown in Table 10.   23 " under "Physiological Aspect X 9 " is shown in Table 11.
, this matrix has an acceptable consistency. That is, under the Physiological Aspect, the weights of "Comfort A 22 " and "Ergonomic Requirement A 23 " are 0.1667 and 0.8333. Under "Psychological Aspect X 10 ", "Functional Aspect X 11 " and "Environmental Aspect X 12 ", there are only "Spiritual Demand A 24 ", "Function Demand A 25 " and "Sustainability A 26 " respectively. That is, the weights of "Spiritual Demand A 24 ", "Function Demand A 25 " and "Sustainability A 26 " relative to their above level are all 1.0000.
The comparison matrix of "Visual Aspect X 1 " and "Emotional Aspect X 2 " under "Aesthetic Attribute Y 1 " is shown in Table 12.

this matrix has an acceptable consistency. That is,
under the Aesthetic Attribute, the weights of "Visual Aspect X 1 " and "Emotional Aspect X 2 " are 0.2500 and 0.7500. The comparison matrix of "Technological Aspect X 3 ", "Use Aspect X 4 ", "Design Aspect X 5 " and "Quality Aspect X 6 " under "Functional Attribute Y 2 " is shown in Table 13.

this matrix has an acceptable
consistency. That is, under the Functional Attribute, the weights of "Technological Aspect X 3 ", "Use Aspect X 4 ", "Design Aspect X 5 " and "Quality Aspect X 6 " are 0.0535, 0.5869, 0.1425 and 0.2172. The comparison matrix of "Value Aspect X 7 " and "Price Aspect X 8 " under "Commercial Attribute Y 3 " is shown in Table 14. under the Commercial Attribute, the weights of "Value Aspect X 7 " and "Price Aspect X 8 " are 0.9000 and 0.1000. The comparison matrix of "Physiological Aspect X 9 ", "Psychological Aspect X 10 ", "Functional Aspect X 11 " and "Environmental Aspect X 12 " under "Human-oriented Attribute Y 4 " is shown in Table 15.

this matrix has an acceptable
consistency. That is, under the Human-oriented Attribute, the weights of "Physiological Aspect X 9 ", "Psychological Aspect X 10 ", "Functional Aspect X 11 " and "Environmental Aspect X 12 " are 0.2151, 0.5131, 0.2151 and 0.0567. The comparison matrix of "Aesthetic Attribute Y 1 ", "Functional Attribute Y 2 ", "Commercial Attribute Y 3 " and "Human-oriented Attribute Y 4 " under "Overall Evaluation of Product Z" is shown in Table 16.
, λ max = 4.0438, CI = 0.0146, CR = 0.01622 < 0.1, this matrix has an acceptable consistency. That is, under the Overall Evaluation of Product, the weights of "Aesthetic Attribute Y 1 ", "Functional Attribute Y 2 ", "Commercial Attribute Y 3 " and "Human-oriented Attribute Y 4 " are 0.0687, 0.3889, 0.1535 and 0.3889. Then, we integrate the calculation results of the weights into the overall product evaluation framework to obtain the overall product evaluation system, as shown in Figure 5.
In the overall product evaluation system, we can find that at the attribute level, the weight order of the four attributes is: Functional Attribute = Human-oriented Attribute > Commercial Attribute > Aesthetic Attribute; at the evaluation index level, the weights of "Spiritual Demand", "Basic Function" and "Social Influence" are 0.1995, 0.1901 and 0.1000, ranking the top 3 in the weight ranking, while the weights of "Material Texture", "Structure and Shape", and "Transportation Cost" are 0.0025, 0.0025, and 0.0014, ranking the last three in the weight ranking. . Overall product evaluation system. Each weight in the system is the weight of each element relative to the element above it. In this study, the weights of elements in the lowest level relative to the element in the highest level, through the hierarchy total ranking, according to the formula: weight of A × weight of X × weight of Y = weight of A relative to Z, can be obtained.
In the overall product evaluation system, we can find that at the attribute level, the weight order of the four attributes is: Functional Attribute = Human-oriented Attribute > Commercial Attribute > Aesthetic Attribute; at the evaluation index level, the weights of "Spiritual Demand", "Basic Function" and "Social Influence" are 0.1995, 0.1901 and 0.1000, ranking the top 3 in the weight ranking, while the weights of "Material Texture", "Structure and Shape", and "Transportation Cost" are 0.0025, 0.0025, and 0.0014, ranking the last three in the weight ranking.

Subjective Product Evaluation System
The overall product evaluation system includes not only subjective evaluation indexes that can be scored by subjective feelings, but also objective evaluation indexes that can only be scored based on the objective information and data of products, rather than the customer's visual impression and subjective feelings of products in the purchase stage. Therefore, in order to establish a reasonable subjective product evaluation system, we need to separate objective and subjective evaluation indexes in the overall product evaluation system, and only retain the subjective evaluation indexes. The weights of the elements in the subjective product evaluation system are further calculated by the weights of the elements in the overall product evaluation system, according to the weight relationships between these elements, as shown in Figure 6. Figure 5. Overall product evaluation system. Each weight in the system is the weight of each element relative to the element above it. In this study, the weights of elements in the lowest level relative to the element in the highest level, through the hierarchy total ranking, according to the formula: weight of A × weight of X × weight of Y = weight of A relative to Z, can be obtained.

Subjective Product Evaluation System
The overall product evaluation system includes not only subjective evaluation indexes that can be scored by subjective feelings, but also objective evaluation indexes that can only be scored based on the objective information and data of products, rather than the customer's visual impression and subjective feelings of products in the purchase stage. Therefore, in order to establish a reasonable subjective product evaluation system, we need to separate objective and subjective evaluation indexes in the overall product evaluation system, and only retain the subjective evaluation indexes. The weights of the elements in the subjective product evaluation system are further calculated by the weights of the elements in the overall product evaluation system, according to the weight relationships between these elements, as shown in Figure 6.
In the subjective product evaluation system, we can find that at the attribute level, the weight order of the four attributes is: Functional Attribute > Human-oriented Attribute > Aesthetic Attribute > Commercial Attribute; at the evaluation index level, the weights of "Spiritual Demand", "Basic Function", and "Function Demand" are 0.2656, 0.2531, and 0.1113, ranking the top 3 in the weight ranking, and the total weight of these three indexes is 0.6300, which is greater than 0.5000, indicating these three indexes to some extent determine the subjective evaluation of a product, as shown in Table 17.
Symmetry 2020, 12, x FOR PEER REVIEW 13 of 25 Figure 6. Subjective product evaluation system. Each weight in the system is the weight of each element relative to the element above it. In this study, the weights of elements in the lowest level relative to the element in the highest level, through the hierarchy total ranking, according to the formula: weight of A × weight of X × weight of Y = weight of A relative to Z, can be obtained.
In the subjective product evaluation system, we can find that at the attribute level, the weight order of the four attributes is: Functional Attribute > Human-oriented Attribute > Aesthetic Attribute > Commercial Attribute; at the evaluation index level, the weights of "Spiritual Demand", "Basic Function", and "Function Demand" are 0.2656, 0.2531, and 0.1113, ranking the top 3 in the weight ranking, and the total weight of these three indexes is 0.6300, which is greater than 0.5000, indicating these three indexes to some extent determine the subjective evaluation of a product, as shown in Table 17.  6. Subjective product evaluation system. Each weight in the system is the weight of each element relative to the element above it. In this study, the weights of elements in the lowest level relative to the element in the highest level, through the hierarchy total ranking, according to the formula: weight of A × weight of X × weight of Y = weight of A relative to Z, can be obtained. Compared with the overall product evaluation system, in the subjective product evaluation system, the weights of "Aesthetic Attribute", "Functional Attribute", and "Human-oriented Attribute" have increased, while the weight of "Commercial Attribute" has decreased, and is less than that of "Aesthetic Attribute". At the same time, "Functional Attribute" occupies the largest weight in two evaluation systems, followed by "Human-oriented Attribute", and the weight of "Aesthetic Attribute" is smaller than "Commercial Attribute" in the overall product evaluation system, but greater than "Commercial Attribute" in the subjective product evaluation system. This means that both in the overall evaluation and subjective evaluation, "Functional Attribute" and "Human-oriented Attribute" of products are very important, and in the overall evaluation, the importance of "Commercial Attribute" is greater than that of "Aesthetic Attribute", but in the subjective evaluation, the importance of "Aesthetic Attribute" is greater than that of "Commercial Attribute", which may be related to the visibility of attributes: subjective evaluation depends largely on visual impressions and related associations caused by visual impressions, and "Aesthetic Attribute" performs better in visibility compared with "Commercial Attribute", so it has a greater importance than "Commercial Attribute" in subjective evaluation.

Experimental Verification
In order to confirm whether the subjective product evaluation system can be used to judge consumer's subjective evaluation of products, we verify it by experiment. The whole experiment includes experimental samples selection and processing, experiment implementation, experimental data processing and analysis, and verification analysis.

Experimental Samples Selection and Processing
In this experiment, we select 10 representative chair products as experimental samples. The feelings brought by visual stimulation can determine the subjective feelings of customers [49,50], and the color arrangements of some experimental samples are not fixed, so we provide experimental samples for the subjects in the form of monochrome pictures. The experimental samples are shown in Figure 7

Experiment Implementation
The experimental method is questionnaire survey. We provide each subject with a quiet and undisturbed questionnaire environment, a paper atlas (see Figure S1) and a paper questionnaire (see Table S1), and there is no time restriction for subjects to complete the questionnaire. The atlas contains pictures and other relevant information (such as material, size, etc.) of each experimental sample. The

Experiment Implementation
The experimental method is questionnaire survey. We provide each subject with a quiet and undisturbed questionnaire environment, a paper atlas (see Figure S1) and a paper questionnaire (see Table S1), and there is no time restriction for subjects to complete the questionnaire. The atlas contains pictures and other relevant information (such as material, size, etc.) of each experimental sample. The questionnaire consists of three parts: the first part is the basic information of the subjects, including gender, age, major background, and education background; the second part is the subjective evaluation of sample chairs, where each sample corresponds to 14 subjective evaluation questions, which are scored by Likert's 5-point scale, as shown in Table 18; the third part is comprehensive evaluation, where the subjects select the "favorite chair" and "least favorite chair" from 10 sample chairs, according to the subjective feelings. A total of 95 subjects participate in this experiment, and the valid questionnaire data come from 91 of them. The gender distribution of these subjects is 49 males (53.8%) and 42 females (46.2%); the ages of these subjects range from 20 to 55 years old, including 75 (82.4%) from 20 to 30 years old, 8 (8.8%) from 31 to 40 years old, and 8 (8.8%) from 41 years old and above; the major backgrounds of these subjects are 41 (45.1%) from design-related majors, 3 (3.3%) from art-related majors, and 47 (51.6%) from other majors; the education backgrounds of these subjects are 25 (27.5%) with college/university education, 60 (65.9%) with master/doctor education, and 6 (6.6%) with other education.  Table 19 shows the subjects' subjective evaluations of each experimental sample. In Table 19, we can get the mean score of each chair under each subjective evaluation question and its standard deviation. For example, Q01: What do you think of the material of this chair? C03 has a mean score of 4.20 and a standard deviation of 0.819 on a 5-point scale; Q06: What do you think of the basic function of this chair? C08 has a mean score of 4.33 and a standard deviation of 0.746 on a 5-point scale. We import the questionnaire data of this experiment into SPSS for reliability and validity analysis. The results show that the Cronbach's Alpha coefficient of the questionnaire is 0.978, greater than 0.900, indicating that the reliability of the questionnaire scale is good; for the analysis items of each chair sample in the questionnaire, the Kaiser-Meyer-Olkin (KMO) test coefficients are all greater than 0.800, and the P values of the Bartlett's sphericity test are all less than 0.010, indicating that there is a correlation between the original data variables of the analysis items, and the data are suitable for factor analysis. In summary, the reliability coefficient of the questionnaire data is greater than 0.900, and the reliability coefficient will not increase significantly after deleting individual analysis item, so the questionnaire data have high reliability and can be used for further analysis.

Effects from Social Factors on Subjective Evaluation
For subjects of different genders, because there are two groups of genders, we use "gender" as the influencing factor to conduct the independent sample t test (T-Test). The differences in subjective evaluation of samples from different genders are shown in Table 20. For subjects of different ages, because there are three groups of ages, we use "age" as the influencing factor to conduct the one-way analysis of variance (ANOVA). The differences in subjective evaluation of samples from different ages are shown in Table 21. For subjects of different major backgrounds, because there are three groups of majors, we use "major background" as the influencing factor to conduct the one-way analysis of variance (ANOVA). The differences in subjective evaluation of samples from different major backgrounds are shown in Table 22. For subjects of different education backgrounds, because there are three groups of education backgrounds, we use "education background" as the influencing factor to conduct the one-way analysis of variance (ANOVA). The differences in subjective evaluation of samples from different education backgrounds are shown in Table 23.  Through the above difference analysis, we can find that differences in social factors do cause differences in some subjective evaluations of some experimental samples. It is embodied as follows: for subjects of different genders, when there are differences in the evaluation of subjective evaluation indexes of experimental samples, the female subjects usually have higher recognitions of these subjective evaluation indexes than the male subjects; for subjects of different ages, when there are differences in the evaluation of subjective evaluation indexes of experimental samples, the subjects aged 20-30 usually have higher recognitions of these subjective evaluation indexes than those aged over 30; for subjects with different major backgrounds, when there are differences in the evaluation of subjective evaluation indexes of experimental samples, the subjects from design-related majors usually have higher recognitions of these subjective evaluation indexes than those from non-design-related majors; for subjects with different education backgrounds, when there are differences in the evaluation of subjective evaluation indexes of experimental samples, the subjects with other education usually have higher recognitions of these subjective evaluation indexes than those with college/university and master/doctor education. However, these differences do not affect the comparison between the mean scores of each subjective evaluation index of each experimental sample.

Verification Analysis
According to the formula: overall subjective evaluation = score of A 1 × weight of A 1 + . . . + score of A n × weight of A n , we can get the overall subjective evaluation of each experimental sample. The overall subjective evaluations of experimental samples are shown in Table 24.
At the same time, we collate and display the survey results of the comprehensive evaluation part of the questionnaire in bar charts, as shown in Figure 8.
In Figure 8a, the statistical result of the subjects' "favorite chair" is that C03 has the highest identification degree, with 23 subjects (25.3%) choosing this option; followed by C08, with 22 subjects (24.2%) choosing this option; and then C06, with 15 subjects (16.5%) choosing this option; and then C10, with 14 subjects (15.4%) choosing this option. C02 has the lowest identification degree, with no one choosing this option; followed by C01, with only 1 subject (1.1%) choosing this option. In Figure 8b, the statistical result of the subjects' "least favorite chair" is that C01 has the highest identification degree, with 27 subjects (29.7%) choosing this option; followed by C02, with 18 subjects (19.8%) choosing this option; and then C07, with 14 subjects (15.4%) choosing this option; and then C09, with 13 subjects (14.3%) choosing this option. C06 has the lowest identification degree, with only 1 subject (1.1%) choosing this option; followed by C03, with 2 subjects (2.2%) choosing this option. At the same time, C04 and C10 have the same identification degree. evaluation of subjective evaluation indexes of experimental samples, the subjects with other education usually have higher recognitions of these subjective evaluation indexes than those with college/university and master/doctor education. However, these differences do not affect the comparison between the mean scores of each subjective evaluation index of each experimental sample.

Verification Analysis
According to the formula: overall subjective evaluation = score of A × weight of A +…+ score of A × weight of A , we can get the overall subjective evaluation of each experimental sample. The overall subjective evaluations of experimental samples are shown in Table 24. At the same time, we collate and display the survey results of the comprehensive evaluation part of the questionnaire in bar charts, as shown in Figure 8. In Figure 8a, the statistical result of the subjects' "favorite chair" is that C03 has the highest identification degree, with 23 subjects (25.3%) choosing this option; followed by C08, with 22 subjects (24.2%) choosing this option; and then C06, with 15 subjects (16.5%) choosing this option; and then C10, with 14 subjects (15.4%) choosing this option. C02 has the lowest identification degree, with no one choosing this option; followed by C01, with only 1 subject (1.1%) choosing this option.
In Figure 8b, the statistical result of the subjects' "least favorite chair" is that C01 has the highest identification degree, with 27 subjects (29.7%) choosing this option; followed by C02, with 18 subjects According to the survey results of comprehensive evaluation, we divide the identification degree of "favorite chair" into ten levels and the identification degree of "least favorite chair" into 9 levels (because, for "least favorite chair", C04 and C10 have the same identification degree). Then, we put the overall subjective evaluation of samples and the identification degree level of samples in comprehensive evaluation in the same coordinate system. In this coordinate system, the horizontal axis is the identification degree level in comprehensive evaluation (the higher the level, the higher the identification degree); the vertical axis is the overall subjective evaluation (the higher the score, the higher the evaluation). Each coordinate point is composed of the identification degree level and the overall subjective evaluation of the experimental sample at this identification degree level, and then these coordinate points together constitute a scatterplot of the relationship between the overall subjective evaluation and comprehensive evaluation of the experimental samples, as shown in Figure 9. axis is the identification degree level in comprehensive evaluation (the higher the level, the higher the identification degree); the vertical axis is the overall subjective evaluation (the higher the score, the higher the evaluation). Each coordinate point is composed of the identification degree level and the overall subjective evaluation of the experimental sample at this identification degree level, and then these coordinate points together constitute a scatterplot of the relationship between the overall subjective evaluation and comprehensive evaluation of the experimental samples, as shown in Figure  9. In Figure 9a, we can find that the overall subjective evaluation is basically positively correlated with the level of "favorite chair", that is, the more popular the chair is in the comprehensive evaluation, the higher the overall subjective evaluation will be. In Figure 9b, we can find that the overall subjective evaluation is basically negatively correlated with the level of "least favorite chair", that is, the less popular the chair is in the comprehensive evaluation, the lower the overall subjective evaluation will be. Therefore, it means that in this study, the overall subjective evaluation of the product can be used to judge the comprehensive evaluation of the product: the higher the overall subjective evaluation, the higher the comprehensive evaluation, and vice versa. The overall subjective evaluations of these products are calculated by the subjective product evaluation system proposed in this study, which means that the subjective product evaluation system can be used to judge the consumer's comprehensive evaluation of the product, that is, the subjective product evaluation system has reliability and validity.

Conclusions
In this study, we comprehensively determine the factors affecting customer feelings in products from four attributes based on KE, take these factors as product evaluation indexes to establish the overall product evaluation system by using AHP, and then obtain the subjective product evaluation system of this study by classifying the evaluation indexes into "subjective evaluation index" and In Figure 9a, we can find that the overall subjective evaluation is basically positively correlated with the level of "favorite chair", that is, the more popular the chair is in the comprehensive evaluation, the higher the overall subjective evaluation will be. In Figure 9b, we can find that the overall subjective evaluation is basically negatively correlated with the level of "least favorite chair", that is, the less popular the chair is in the comprehensive evaluation, the lower the overall subjective evaluation will be. Therefore, it means that in this study, the overall subjective evaluation of the product can be used to judge the comprehensive evaluation of the product: the higher the overall subjective evaluation, the higher the comprehensive evaluation, and vice versa. The overall subjective evaluations of these products are calculated by the subjective product evaluation system proposed in this study, which means that the subjective product evaluation system can be used to judge the consumer's comprehensive evaluation of the product, that is, the subjective product evaluation system has reliability and validity.

Conclusions
In this study, we comprehensively determine the factors affecting customer feelings in products from four attributes based on KE, take these factors as product evaluation indexes to establish the overall product evaluation system by using AHP, and then obtain the subjective product evaluation system of this study by classifying the evaluation indexes into "subjective evaluation index" and "objective evaluation index", eliminating the objective evaluation indexes, and retaining the subjective evaluation indexes. In the subjective product evaluation system, at the attribute level, "Functional Attribute" and "Human-oriented Attribute" are more important than "Aesthetic Attribute" and "Commercial Attribute"; at the evaluation index level, "Spiritual Demand", "Basic Function" and "Function Demand" are more important than other indexes.
Through a verification experiment with 10 representative chair products as experimental objects, it can be concluded the subjective product evaluation system based on KE and AHP proposed in this study has reliability and validity. It realizes a symmetry between subjective product evaluation and comprehensive product evaluation, making it possible to complete comprehensive evaluation of a product through customer subjective feelings without objective information of this product.
This subjective product evaluation system can be used to select products with higher evaluation, and then further analyze the characteristics of these products at different levels of this system, so as to improve existing products or develop new products that meet people's purchase wishes; it can also be used to evaluate the classic products in different periods of society, and then summarize the characteristics of classic products and the characteristics of social development in different periods; it can also be used to study consumer's psychology, explore consumer's preferences for various products from different social backgrounds, and then summarize the laws for targeted product promotion. In the future, we will strive to further simplify this subjective product evaluation system and develop a corresponding product evaluation software for rapidly evaluating products.
Author Contributions: Y.Z. designed and performed the experiments, analyzed the experimental data, and wrote the manuscript. Z.W. designed the study, provided the research materials, and gave some advice on experiments design and manuscript writing. All authors have read and agreed to the published version of the manuscript.

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