Tourism consumer demand analysis method and system based on improved KANO model

Abstract

The present invention provides a method and system for analyzing tourism consumer demand based on an improved KANO model, relating to the field of online tourism consumer demand analysis. The method first utilizes a sentiment dictionary based on online tourism review data to obtain fine-grained sentiment values ​​for product attributes. Then, an improved KANO model is constructed based on these sentiment values. The improved KANO model includes: utilizing positive and negative sentiment coefficients within the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance; and introducing product attribute demand weights into the KANO model. Finally, tourism consumer demand analysis is performed based on the improved KANO model. The present invention enables a more refined classification of consumer demands from online tourism review data, resulting in more accurate tourism consumer demand analysis results compared to existing technologies.

Description

Technical Field

[0001] The present invention relates to the technical field of online tourism consumer demand analysis, and in particular to a tourism consumer demand analysis method and system based on an improved KANO model. Background Art

[0002] Analyzing consumer demand helps companies gather and understand the product features and requirements of different consumers, thereby providing decision-making support for product improvements and enhancements. With the development of the online tourism industry, more and more consumers are posting personal reviews on online travel platforms after their trips. In recent years, some scholars have demonstrated that online reviews can serve as an effective data source for demand analysis, and that the sentiment contained in online reviews can effectively represent the demand preferences of travel consumers.

[0003] Currently, research on tourism consumer demand analysis based on online reviews primarily focuses on analyzing the overall sentiment of the text, identifying and ranking key consumer needs, and identifying factors influencing consumer satisfaction. Furthermore, research methods for analyzing consumer demand include principal component analysis, the KANO model, and the coupling coordination model. Compared to other methods, the KANO model is more widely used because it can describe the nonlinear relationship between different product attributes and consumer satisfaction, allowing for classification and analysis of different consumer demand types.

[0004] However, tourism consumers often have different demands for various attributes of tourism products. Existing methods, which only analyze the overall sentiment preferences of text, cannot accurately reflect and analyze consumers' demand preferences for each product attribute. Secondly, existing KANO models and methods based on online reviews have difficulty measuring changes in consumer satisfaction when product performance changes, and they also cannot measure the importance consumers place on different product attributes. Therefore, existing tourism consumer demand analysis suffers from low accuracy. Summary of the Invention

[0005] (1) Technical problems solved

[0006] In view of the shortcomings of the existing technology, the present invention provides a tourism consumer demand analysis method and system based on the improved KANO model, which solves the problem of low accuracy in the existing tourism consumer demand analysis.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] In a first aspect, the present invention first proposes a tourism consumer demand analysis method based on an improved KANO model, the method comprising:

[0010] Based on pre-acquired online travel review data, we use sentiment dictionaries to obtain fine-grained sentiment values ​​of travel consumers on product attributes.

[0011] Constructing an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model;

[0012] Tourism consumer demand analysis based on the improved KANO model.

[0013] Preferably, the method of obtaining fine-grained sentiment values ​​of tourism consumers on product attributes using a sentiment dictionary based on pre-acquired online tourism review data includes:

[0014] S11, pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols;

[0015] S12, extracting candidate product attribute feature words from the pre-processed online review data;

[0016] S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words;

[0017] S14, calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determining the seed sentiment words under each attribute based on the co-occurrence, extracting all adjectives in the online reviews as candidate sentiment words, calculating the similarity between the candidate sentiment words and the seed sentiment words using a point mutual information algorithm, and adding the candidate sentiment words whose similarity meets the threshold to the basic sentiment dictionary to construct a domain sentiment dictionary;

[0018] S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

[0019] Preferably, constructing an improved KANO model based on the sentiment value includes:

[0020] S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value;

[0021] S22. Use the random forest model to determine the attribute demand weights of the product;

[0022] S23, determining positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measuring the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference;

[0023] S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

[0024] Preferably, the basic sentiment dictionary includes: HowNet sentiment dictionary.

[0025] Preferably, the nonlinear relationship between consumer satisfaction and product attribute performance measured based on the positive and negative sentiment coefficients includes: a,pos,high , P a,neg,low Two indicators are used to measure the continuity relationship between a product's attribute function and consumer satisfaction; among them,

[0026] P a,pos,high =P a,pos,4 +P a,pos,5

[0027] P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1

[0028] Among them, P a,pos,high , P a,neg,low They represent the probability that a high score for product attribute a will result in a high total score and the probability that a low score for product attribute a will result in a low total score; P a,pos,s , P a,neg,s They represent the probability of generating a total score s when a positive evaluation and a negative evaluation are generated for attribute a, s = 1, 2, 3, 4, 5 respectively.

[0029] In a second aspect, the present invention further proposes a tourism consumer demand analysis system based on an improved KANO model, the system comprising:

[0030] The fine-grained sentiment value acquisition module is used to obtain the fine-grained sentiment values ​​of tourism consumers on product attributes based on the pre-acquired online tourism review data using the sentiment dictionary;

[0031] An improved KANO model construction module is used to construct an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model;

[0032] The tourism consumer demand analysis result output module is used to analyze tourism consumer demand based on the improved KANO model and output the analysis results.

[0033] Preferably, the fine-grained sentiment value acquisition module acquires the fine-grained sentiment values ​​of tourism consumers on product attributes using a sentiment dictionary based on pre-acquired online tourism review data, including:

[0034] S11, pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols;

[0035] S12, extracting candidate product attribute feature words from the pre-processed online review data;

[0036] S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words;

[0037] S14, calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determining the seed sentiment words under each attribute based on the co-occurrence, extracting all adjectives in the online reviews as candidate sentiment words, calculating the similarity between the candidate sentiment words and the seed sentiment words using a point mutual information algorithm, and adding the candidate sentiment words whose similarity meets the threshold to the basic sentiment dictionary to construct a domain sentiment dictionary;

[0038] S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

[0039] Preferably, the improved KANO model construction module constructs the improved KANO model based on the sentiment value, including:

[0040] S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value;

[0041] S22. Use the random forest model to determine the attribute demand weights of the product;

[0042] S23, determining positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measuring the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference;

[0043] S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

[0044] Preferably, the basic sentiment dictionary includes: HowNet sentiment dictionary.

[0045] Preferably, the nonlinear relationship between consumer satisfaction and product attribute performance measured based on the positive and negative sentiment coefficients includes: a,pos,high , P a,neg,low Two indicators are used to measure the continuity relationship between a product's attribute function and consumer satisfaction; among them,

[0046] P a,pos,high =P a,pos,4 +P a,pos,5

[0047] P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1

[0048] Among them, P a,pos,high , P a,neg,low They represent the probability that a high score for product attribute a will result in a high total score and the probability that a low score for product attribute a will result in a low total score; P a,pos,s , P a,neg,s They represent the probability of generating a total score s when a positive evaluation and a negative evaluation are generated for attribute a, s = 1, 2, 3, 4, 5 respectively.

[0049] (3) Beneficial effects

[0050] The present invention provides a tourism consumer demand analysis method and system based on an improved KANO model. Compared with the existing technology, it has the following advantages:

[0051] 1. The present invention first uses a sentiment dictionary based on online travel review data to obtain fine-grained sentiment values ​​for product attributes. Then, based on the sentiment values, an improved KANO model is constructed. The improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model. Finally, tourism consumer demand analysis is performed based on the improved KANO model. The present invention mines fine-grained consumer sentiment from online travel review data and uses the improved KANO model to introduce product attribute demand weights to perform a more refined division of demand. Compared with existing technologies, this method can obtain more accurate tourism consumer demand analysis results.

[0052] 2. The present invention identifies product attributes more comprehensively by recognizing fixed multi-word expressions in consumers' online reviews, and then constructs a tourism-related sentiment dictionary based on the existing basic sentiment dictionary. The tourism-related sentiment dictionary can be used to mine the fine-grained emotions of tourism consumers from online reviews, and then characterize consumers' demand preferences for various attributes. It can more accurately analyze consumer needs, thereby providing effective data support for the functional design of various product attribute dimensions.

[0053] 3. The present invention uses an improved KANO model to realize tourism consumer demand analysis, and the improved KANO model includes two levels of improvement to the traditional KANO model. The first level of improvement constructs a KANO model based on emotional preference. Through this model, the positive and negative emotional coefficients of each attribute in the online review data can be mined, and the changes in consumer satisfaction with product performance can be quantified; the second level of improvement is to add the demand weights of each attribute of the product on the basis of the KANO model based on emotional preference, so as to measure the importance of different product attributes, thereby quantifying the different levels of importance consumers attach to each product attribute, and more accurately analyzing consumer needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 This is a flow chart of a tourism consumer demand analysis method based on an improved KANO model of the present invention;

[0056] Figure 2 Flowchart of a fine-grained sentiment analysis method according to an embodiment of the present invention;

[0057] Figure 3 This is a flow chart of performing tourism consumer demand analysis based on the improved KANO model in an embodiment of the present invention;

[0058] Figure 4 The KANO model based on emotional preference in the embodiment of the present invention;

[0059] Figure 5 This is the improved KANO model in the embodiment of the present invention. DETAILED DESCRIPTION

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] The embodiments of the present application provide a tourism consumer demand analysis method and system based on an improved KANO model, thereby solving the problem of low accuracy in existing tourism consumer demand analysis and achieving the purpose of providing effective decision support for product design of online travel companies.

[0062] The technical solution in the embodiments of the present application is to solve the above technical problems, and the overall idea is as follows:

[0063] Because consumers’ demand preferences for various product attributes are not taken into account, and the existing KANO model does not measure the importance of different product attributes, which makes it impossible to reflect the different levels of importance consumers place on various product attributes, the existing tourism consumer demand analysis technology based on online reviews has the problem of low accuracy. In order to solve the above problems, the technical solution of this application constructs a tourism consumer demand analysis model based on fine-grained emotions and an improved KANO model. Through feature extraction algorithms and sentiment analysis algorithms, the different attributes of tourism products and fine-grained consumer sentiment information for each attribute are identified. Machine learning algorithms and statistical analysis methods are applied to mine consumers’ positive and negative preference information in online reviews to measure the nonlinear relationship between product attribute performance and consumer satisfaction, and the improved traditional KANO model is used to identify and classify different consumer demands. The tourism consumer demand analysis model constructed by this application effectively converts online review data into product demand classification, while overcoming the shortcoming of the KANO model that cannot measure demand weights, and can provide effective decision support for product design of online travel companies.

[0064] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0065] Example 1:

[0066] In the first aspect, the present invention first proposes a tourism consumer demand analysis method based on the improved KANO model. Figure 1 , the method comprising:

[0067] S1. Based on pre-acquired online travel review data, the sentiment dictionary is used to obtain fine-grained sentiment values ​​of travel consumers on product attributes;

[0068] S2. constructing an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model;

[0069] S3. Analyze tourism consumer demand based on the improved KANO model.

[0070] As can be seen, this embodiment first uses a sentiment dictionary based on online travel review data to obtain fine-grained sentiment values ​​for product attributes; then, based on the sentiment values, an improved KANO model is constructed. The improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model; finally, tourism consumer demand analysis is performed based on the improved KANO model. The present invention mines fine-grained consumer sentiment from online travel review data and uses the improved KANO model to introduce product attribute demand weights to perform a more refined division of demand. Compared with existing technologies, more accurate tourism consumer demand analysis results can be obtained.

[0071] The following is combined with Figure 1-5 , and explanations of specific steps S1-S3 are provided to describe in detail the implementation process of an embodiment of the present invention.

[0072] S1. Based on the pre-acquired online travel review data, the sentiment dictionary is used to obtain the fine-grained sentiment values ​​of travel consumers towards product attributes.

[0073] Since tourism products have many attributes, and only a few consumers care about some of them, machine learning methods are too expensive and inaccurate to train models. Therefore, this embodiment proposes a fine-grained sentiment analysis method for product attributes based on a sentiment dictionary. This method aims to identify different attributes of online tourism products and determine the fine-grained sentiment of tourism consumers towards each attribute. Figure 2 As shown, the specific steps of this method are as follows:

[0074] S11 , pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols.

[0075] This embodiment first crawls online reviews containing personal opinions from consumers on online travel service platforms and related websites. This data is then preprocessed, including removing invalid comments such as short and irrelevant ones, including advertisements. Subsequently, the removed data is segmented and each segmented word is tagged with a part-of-speech tag. During the segmentation process, all stop words, including common conjunctions and prepositions, are removed to ensure the accuracy of the segmentation results. Furthermore, the method of this embodiment converts special symbols such as emoticons into text to ensure that emotional expressions are not overlooked.

[0076] S12: Extracting candidate product attribute feature words from the pre-processed online review data.

[0077] Product attribute feature words often appear frequently in online reviews as nouns. In addition to single nouns, product attribute feature words may also include fixed multi-word expressions. Fixed multi-word expressions contain several segmented words, but they can be considered as a complete word, which makes them difficult to accurately identify during the segmentation process. Therefore, this embodiment proposes a new method for extracting product attribute feature words that considers fixed multi-word expressions.

[0078] This embodiment selects point mutual information and left-right information entropy to measure the relevance between words to identify fixed multi-word expressions. The definition of point mutual information is shown in formula (1):

[0079]

[0080] Among them, P(w i ,w j ) represents the word w i and w j The probability of co-occurrence, and P(w i ) and P(w j ) represent the word w i and w j The probability of appearing alone. The larger the mutual information value of a point, the greater the correlation between the two.

[0081] The definition of left and right information entropy is shown in formula (2) (3):

[0082] IE left(X)=∑P(PreX|X)×log2P(PreX|X) (2)

[0083] IE right (X)=∑P(XPost|X)×log2P(XPost|X) (3)

[0084] Here, X represents the phrase being analyzed, Pre and Post represent the individual words adjacent to the left and right of X, respectively. P(PreX|X) represents the conditional probability that the word to the left of X is Pre when X appears; P(XPost|X) represents the conditional probability that the word to the right of X is Pre when X appears. The left and right information entropies can be used to determine whether a word is a boundary word. The larger the left and right information entropies, the more likely the word is an independent word. This paper selects the smaller value between the left and right information entropies and adds it to the point mutual information value to represent the correlation between words. The higher the final value, the more likely it is a fixed multi-word expression. This embodiment uses the BNP (BaseNounPhrase) structure to filter all identified fixed multi-word expressions. Thus, by performing frequent item set identification on all nouns in the word segmentation results and the filtered fixed multi-word expressions, candidate product attribute feature words in online reviews can be preliminarily extracted.

[0085] S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words.

[0086] Because consumers typically use different words to represent the same attribute, for example, the "accommodation experience" attribute in a tourism product may include words such as "hotel" and "guesthouse." Therefore, this embodiment requires filtering and categorizing the extracted product attribute feature words according to their attributes. Specifically, this embodiment selects the LDA model for product attribute identification. The LDA model is a widely used bag-of-words model that is widely used to extract text topics. This embodiment uses the perplexity index to determine the optimal number of topics, and supplements and adjusts them through interviews with tourism product design experts, thereby identifying all product attributes.

[0087] Considering that product attribute feature words belonging to the same attribute have a high semantic similarity, this embodiment classifies all extracted candidate product attribute feature words by manually setting seed attribute words and calculating semantic similarity. Semantic similarity refers to the possibility that two words in a text can be replaced with each other without changing the original meaning of the sentence. It is often used to characterize the similarity of meaning between words. This embodiment uses the Word2vec algorithm to convert the above candidate product attribute feature words into word vectors and calculates their similarity through cosine similarity. The cosine similarity calculation formula for n-dimensional vectors is as follows:

[0088]

[0089] Where x and y represent the two vectors corresponding to the candidate product attribute feature word and the seed attribute word, respectively, and Sim(x, y) represents the similarity between the two. The closer the Sim(x, y) value is to 1, the higher the similarity. Thus, by comparing the similarities, the extracted candidate product attribute feature words are screened and assigned to the corresponding attributes. In this embodiment, candidate product attribute feature words with a similarity greater than 0.75 are classified as the same attribute.

[0090] S14. Calculate the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determine the seed sentiment words under each attribute based on the co-occurrence, extract all adjectives in the online reviews as candidate sentiment words, calculate the similarity between the candidate sentiment words and the seed sentiment words with the help of the point mutual information algorithm, and classify the candidate sentiment words whose similarity meets the threshold into the basic sentiment dictionary to construct a domain sentiment dictionary.

[0091] Online reviews contain consumers' sentimental preferences for various product attributes. This example utilizes a fine-grained sentiment analysis method based on a sentiment lexicon to determine consumer preferences for different attributes within each review. Sentiment in online reviews is primarily expressed by the sentiment words they contain. Currently, many research groups draw on existing sentiment lexicons, such as HowNet's Sentiment Dictionary and the Dalian University of Technology's Sentiment Dictionary. However, these sentiment dictionaries contain a limited number of sentiment words. To identify sentiment words specific to the online travel industry and improve sentiment analysis accuracy, this example constructs a domain sentiment lexicon based on the basic sentiment lexicon and then proposes a fine-grained sentiment analysis method based on this domain sentiment lexicon.

[0092] This paper uses the HowNet sentiment dictionary as the basic sentiment dictionary and determines the seed sentiment words for each attribute by calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words classified as above. The co-occurrence calculation formula is as follows:

[0093]

[0094] Where S and A represent sentiment words and product attribute feature words, respectively; W(S, A) represents the co-occurrence between sentiment words and product attribute feature words; F(S, A) represents the number of times both appear together in an online review; and F(X) and F(A) represent the number of times each appears alone in a review. Sentiment words with a high co-occurrence (in this embodiment, a co-occurrence greater than 0.3 is defined as a high co-occurrence) are selected as seed sentiment words. To comprehensively identify sentiment words, this embodiment extracts all adjectives in online reviews as candidate sentiment words. Using the pointwise mutual information algorithm, the similarity between these words and the seed sentiment words is calculated. Words with a high similarity (greater than 0.75) are then assigned to a sentiment dictionary and merged with the basic sentiment dictionary to form a domain-specific sentiment dictionary suitable for analyzing online travel consumer review needs.

[0095] S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

[0096] Sentiment preference identification is performed based on the aforementioned domain sentiment dictionary. The sentiment value of the product attribute feature word is quantified by searching the text before and after the attribute word for sentiment words, degree adverbs, and negative adverbs. When a sentiment word is positive, it indicates that consumers are relatively satisfied with the product's performance in relation to that attribute. Degree adverbs complement sentiment words, such as "extremely" increasing sentiment intensity and "slightly" decreasing it. Negative adverbs can reverse sentiment polarity. This paper quantifies sentiment intensity on a scale of 1-5, where 1 indicates extreme dissatisfaction, 3 indicates moderate sentiment, 5 indicates extreme satisfaction, and 2 and 4 fall in between.

[0097] S2. Constructing an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing the product attribute demand weight into the KANO model.

[0098] An online review will involve consumers' fine-grained emotions towards various product attributes, and this kind of emotional preference information can be used to characterize consumer demand. This kind of fragmented demand information needs to be transformed into reasonable demand classification through demand integration analysis to guide product design and improvement. Based on this, this embodiment proposes an improved KANO model based on the traditional KANO model for tourism consumer demand analysis. The specific process is as follows: Figure 3The improved KANO model includes two levels of improvement: the first level of improvement is: in the traditional KANO model, the nonlinear relationship between consumer satisfaction and product performance is measured by the positive and negative sentiment coefficients in online reviews; the second level of improvement is: the attribute demand weight of the product is introduced into the traditional KANO model.

[0099] S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value.

[0100] When consumers post online reviews, they do not comment on all attributes of a product. Therefore, in this embodiment, we use the TOPSIS missing value estimation method to estimate missing sentiment for product attributes that consumers have not commented on based on large-scale data.

[0101] When the total product rating is known, the missing sentiment tendency is expressed in the form of conditional probability. The calculation formula is as follows:

[0102] (1) The upper and lower bounds of the probability of missing a sentiment score of 5:

[0103]

[0104]

[0105] (2) The upper and lower bounds of the probability of missing a sentiment score of 4:

[0106]

[0107]

[0108] The probability of missing sentiment scores in other situations can be deduced in the same way. u ) represents the number of times a given total score is y, N(x u,a =n,y u ), N(x u,a =-,y u ) represent the number of times a consumer gave a score of n and did not give a score for attribute a. Therefore, we can derive the probability range of the missing value taking different values ​​when the total score is determined, which is The maximum probability range is [1,1], and the minimum probability range is [0,0]. The distance from the maximum probability range and the minimum probability range is as follows:

[0109]

[0110]

[0111] Among them, P + , P - Represents the distance from the maximum probability range and the minimum probability range respectively. The probability range is converted into a specific probability value using the TOPSIS method. The calculation formula is as follows:

[0112]

[0113] Therefore, when the total score is determined, the probability density P of the missing sentiment value can be obtained, and the missing sentiment value can be filled according to the probability value to obtain the customer's sentiment value for all attributes of the product.

[0114] S22. Use the random forest model to determine the attribute demand weights of the product.

[0115] When consumers comment on products, the importance of different attributes varies. However, the traditional KANO model does not consider the weights between different attributes. Therefore, this embodiment proposes a weight determination method based on random forest.

[0116] When a random forest generates a tree, its internal nodes will select an attribute from the candidate attributes as the splitting attribute of the node when splitting. Using the above characteristics, one of the attributes is randomly replaced in each iteration. The change in model accuracy after the replacement is used to predict the degree of influence of the attribute on the overall result, that is, the attribute demand weight of the product.

[0117] S23. Determine positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measure the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference.

[0118] Since the consumer's emotional value and overall rating of a product attribute are both discrete data, it is difficult to measure the continuity relationship between the two. Therefore, this embodiment analyzes by constructing a discrete correlation between the two. When consumers give high scores to a certain attribute and the overall rating is also high, it means that when the attribute is satisfied, the consumer satisfaction will increase; conversely, when a certain attribute is scored low and the overall rating is also low, it means that when the attribute cannot be satisfied, the overall satisfaction will decrease. Based on the above ideas, this embodiment proposes a KANO model based on emotional preference, such as Figure 4 As shown. a,pos,high , P a,neg,low They respectively represent the probability that a high score for product attribute a (a score of 4 or 5 points) will result in a high total score, and the probability that a low score for product attribute a (a score of 1, 2 or 3 points) will result in a low total score. The above two indicators are used to measure the continuity relationship between a certain attribute function of a product and consumer satisfaction.

[0119] In this example, scores of 4 and 5 are generally considered to be satisfactory to consumers, while scores of 3 and below are considered to be unsatisfactory. The following calculation formula is thus given:

[0120] P a,pos,high =P a,pos,4 +P a,pos,5

[0121] P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1

[0122] Among them, P a,pos,s , P a,neg,s They represent the probability of generating a total score s (s = 1, 2, 3, 4, 5) when a positive evaluation (score of 4 or 5) and a negative evaluation (score of 1 or 2 or 3) are generated for attribute a. a,pos,high , P a,neg,low According to the numerical value of , consumer demand can be divided into four categories, namely, attractive demand, indifferent demand, necessary demand, and expected demand.

[0123] S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

[0124] Subsequently, this embodiment introduces the weight of each attribute requirement of the product into the above discrete KANO model to construct a three-dimensional space, that is, to perform a second-level improvement on the traditional KANO model. The final improved KANO model obtained after the second-level improvement is as follows: Figure 5 shown.

[0125] S3. Analyze tourism consumer demand based on the improved KANO model.

[0126] After the above steps S1-S2, we constructed an improved KANO model, which can be used to analyze tourism consumer demand through consumer online reviews. Figure 5 The specific weights of consumers' demands for different product attributes and P are given in a,neg,low Value and P a,pos,high The relationship between the values ​​is as follows:

[0127] (1) Core functional requirements: When P a,neg,low The higher the value, the a,pos,high If the value is low, the attribute is considered a basic requirement. This type of requirement is generally considered the most basic function that the product needs to provide. If the attribute weight is high, it can be considered the most core functional requirement of the product.

[0128] (2) Ensure quality requirements: When Pa,neg,low The higher the value, the a,pos,high When the value is low, the attribute is a basic requirement. When the attribute weight is low, it means that the attribute is not the most core function of the product, but it is indispensable and the performance quality of the attribute must be guaranteed.

[0129] (3) Prioritize demand: When P a,pos,high , P a,neg,low When both values ​​are high, the attribute is a desired requirement. Improving product performance with this attribute can significantly increase consumer satisfaction. When this attribute has a high weight, companies should prioritize improving it.

[0130] (4) Suboptimal improvement demand: When P a,pos,high , P a,neg,low When the values ​​are both high, the attribute is a desired need. When the attribute weight is low, it means that most consumers do not currently value it. Therefore, companies need to ensure investment in this attribute, but they should give it a low priority.

[0131] (5) Prioritize innovation needs: When P a,pos,high The higher the value, the a,neg,low When the value is low, this attribute is considered a compelling need. When the weight of this product attribute is high, it indicates that improving this need can significantly increase consumer satisfaction. However, since insufficient performance of this attribute will not cause a decrease in satisfaction, it is not given the highest priority and requires companies to increase their innovation efforts.

[0132] (6) Implicit charm needs: When P a,pos,high The higher the value, the a,neg,low When the value is low, the attribute is a charm demand. When the weight of the attribute is low, it means that it is less valued by consumers and belongs to the implicit charm demand.

[0133] (7) When P a,pos,high , P a,neg,low When the values ​​are all low, the attribute is undifferentiated. When the attribute weight is high, even if the performance is improved, it will not significantly increase consumer satisfaction, so companies should avoid over-investment.

[0134] (8) When P a,pos,high , P a,neg,low When the values ​​are all low, the attribute is an indifferent demand. When the attribute weight is low, it means that its performance is basically irrelevant to the overall evaluation of the product and is an irrelevant demand.

[0135] Ultimately, tourism product producers can judge tourism consumers' demand and attention to various product attributes based on the above correspondence. Based on the demand analysis results, they can provide effective decision-making support for product design of online tourism companies.

[0136] At this point, the entire process of the tourism consumer demand analysis method based on the improved KANO model of the present invention is completed.

[0137] Example 2:

[0138] In a second aspect, the present invention further provides a tourism consumer demand analysis system based on an improved KANO model, the system comprising:

[0139] The fine-grained sentiment value acquisition module is used to obtain the fine-grained sentiment values ​​of tourism consumers on product attributes based on the pre-acquired online tourism review data using the sentiment dictionary;

[0140] An improved KANO model construction module is used to construct an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model;

[0141] The tourism consumer demand analysis result output module is used to analyze tourism consumer demand based on the improved KANO model and output the analysis results.

[0142] Optionally, the fine-grained sentiment value acquisition module acquires the fine-grained sentiment values ​​of tourism consumers toward product attributes using a sentiment dictionary based on pre-acquired online tourism review data, including:

[0143] S11, pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols;

[0144] S12, extracting candidate product attribute feature words from the pre-processed online review data;

[0145] S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words;

[0146] S14, calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determining the seed sentiment words under each attribute based on the co-occurrence, extracting all adjectives in the online reviews as candidate sentiment words, calculating the similarity between the candidate sentiment words and the seed sentiment words using a point mutual information algorithm, and adding the candidate sentiment words whose similarity meets the threshold to the basic sentiment dictionary to construct a domain sentiment dictionary;

[0147] S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

[0148] Optionally, the improved KANO model construction module constructs an improved KANO model based on the sentiment value, including:

[0149] S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value;

[0150] S22. Use the random forest model to determine the attribute demand weights of the product;

[0151] S23, determining positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measuring the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference;

[0152] S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

[0153] Optionally, the basic sentiment dictionary includes: HowNet sentiment dictionary.

[0154] Optionally, the nonlinear relationship between consumer satisfaction and product attribute performance measured based on the positive and negative sentiment coefficients includes: a,pos,high , P a,neg,low Two indicators are used to measure the continuity relationship between a product's attribute function and consumer satisfaction; among them,

[0155] P a,pos,high =P a,pos,4 +P a,pos,5

[0156] P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1

[0157] Among them, P a,pos,high , P a,neg,low They represent the probability that a high score for product attribute a will result in a high total score and the probability that a low score for product attribute a will result in a low total score; P a,pos,s , P a,neg,s They represent the probability of generating a total score s when a positive evaluation and a negative evaluation are generated for attribute a, s = 1, 2, 3, 4, 5 respectively.

[0158] It can be understood that the tourism consumer demand analysis system based on the improved KANO model provided in the embodiment of the present invention corresponds to the tourism consumer demand analysis method based on the improved KANO model mentioned above. The explanation, examples, beneficial effects and other parts of its relevant contents can refer to the corresponding contents in the tourism consumer demand analysis method based on the improved KANO model, and will not be repeated here.

[0159] In summary, compared with the existing technology, the present invention has the following beneficial effects:

[0160] 1. The present invention first uses a sentiment dictionary based on online travel review data to obtain fine-grained sentiment values ​​for product attributes. Then, based on the sentiment values, an improved KANO model is constructed. The improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model. Finally, tourism consumer demand analysis is performed based on the improved KANO model. The present invention mines fine-grained consumer sentiment from online travel review data and uses the improved KANO model to introduce product attribute demand weights to perform a more refined division of demand. Compared with existing technologies, this method can obtain more accurate tourism consumer demand analysis results.

[0161] 2. The present invention identifies product attributes more comprehensively by recognizing fixed multi-word expressions in consumers' online reviews, and then constructs a tourism-related sentiment dictionary based on the existing basic sentiment dictionary. The tourism-related sentiment dictionary can be used to mine the fine-grained emotions of tourism consumers from online reviews, and then characterize consumers' demand preferences for various attributes. It can more accurately analyze consumer needs, thereby providing effective data support for the functional design of various product attribute dimensions.

[0162] 3. The present invention uses an improved KANO model to realize tourism consumer demand analysis, and the improved KANO model includes two levels of improvement to the traditional KANO model. The first level of improvement constructs a KANO model based on emotional preference. Through this model, the positive and negative emotional coefficients of each attribute in the online review data can be mined, and the changes in consumer satisfaction with product performance can be quantified; the second level of improvement is to add the demand weights of each attribute of the product on the basis of the KANO model based on emotional preference, so as to measure the importance of different product attributes, thereby quantifying the different levels of importance consumers attach to each product attribute, and more accurately analyzing consumer needs.

[0163] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0164] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A tourism consumer demand analysis method based on the improved KANO model, characterized by: The method comprises: Based on pre-acquired online travel review data, we use sentiment dictionaries to obtain fine-grained sentiment values ​​of travel consumers on product attributes. Constructing an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model; Analyze tourism consumer demand based on the improved KANO model; The step of constructing an improved KANO model based on the sentiment value includes: S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value; S22. Use the random forest model to determine the attribute demand weights of the product; S23, determining positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measuring the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference; S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

2. The method according to claim 1, wherein The method of obtaining fine-grained sentiment values ​​of tourism consumers on product attributes using sentiment dictionaries based on pre-acquired online tourism review data includes: S11, pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols; S12, extracting candidate product attribute feature words from the pre-processed online review data; S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words; S14, calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determining the seed sentiment words under each attribute based on the co-occurrence, extracting all adjectives in the online reviews as candidate sentiment words, calculating the similarity between the candidate sentiment words and the seed sentiment words using a point mutual information algorithm, and adding the candidate sentiment words whose similarity meets the threshold to the basic sentiment dictionary to construct a domain sentiment dictionary; S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

3. The method according to claim 2, wherein The basic sentiment dictionary includes: HowNet sentiment dictionary.

4. The method according to claim 1, wherein The nonlinear relationship between consumer satisfaction and product attribute performance measured based on the positive and negative sentiment coefficients includes: a,pos,high , P a,neg,low Two indicators are used to measure the continuity relationship between a product's attribute function and consumer satisfaction; among them, P a,pos,high =P a,pos,4 +P a,pos,5 P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1 Among them, P a,pos,high , P a,neg,low They represent the probability that a high score for product attribute a will result in a high total score and the probability that a low score for product attribute a will result in a low total score; P a,pos,s , P a,neg,s They represent the probability of generating a total score s when a positive evaluation and a negative evaluation are generated for attribute a, s = 1, 2, 3, 4, 5 respectively.

5. A tourism consumer demand analysis system based on the improved KANO model, characterized by: The system comprises: The fine-grained sentiment value acquisition module is used to obtain the fine-grained sentiment values ​​of tourism consumers on product attributes based on the pre-acquired online tourism review data using the sentiment dictionary; An improved KANO model construction module is used to construct an improved KANO model based on the sentiment value; the improved KANO model includes: using positive and negative sentiment coefficients in the KANO model to measure the nonlinear relationship between consumer satisfaction and product attribute performance, and introducing product attribute demand weights into the KANO model; Tourism consumer demand analysis result output module, used to analyze tourism consumer demand based on the improved KANO model and output the analysis results; The improved KANO model construction module constructs an improved KANO model based on the sentiment value, including: S21. Based on the sentiment value, use TOPSIS to estimate the missing sentiment value to obtain an estimated value of the missing sentiment value; S22. Use the random forest model to determine the attribute demand weights of the product; S23, determining positive and negative sentiment coefficients based on the sentiment value and the estimated value of the missing sentiment value, and measuring the nonlinear relationship between consumer satisfaction and product attribute performance based on the positive and negative sentiment coefficients in the KANO model to construct a KANO model based on sentiment preference; S24. Introducing the attribute demand weights of the product into the KANO model based on emotional preference to construct an improved KANO model.

6. The system according to claim 5, wherein: The fine-grained sentiment value acquisition module uses a sentiment dictionary to acquire fine-grained sentiment values ​​of tourism consumers on product attributes based on pre-acquired online travel review data, including: S11, pre-processing the pre-acquired online review data, wherein the pre-processing includes deleting invalid reviews, word segmentation, and text conversion of special symbols; S12, extracting candidate product attribute feature words from the pre-processed online review data; S13. Extract all product attributes based on the LDA model, set seed attribute words for each product attribute, calculate the similarity between the candidate product attribute feature words and the seed attribute words, and classify all candidate product attribute feature words whose similarity exceeds a preset threshold into the same category of product attribute feature words; S14, calculating the co-occurrence between the sentiment words in the basic sentiment dictionary and the product attribute feature words, determining the seed sentiment words under each attribute based on the co-occurrence, extracting all adjectives in the online reviews as candidate sentiment words, calculating the similarity between the candidate sentiment words and the seed sentiment words using a point mutual information algorithm, and adding the candidate sentiment words whose similarity meets the threshold to the basic sentiment dictionary to construct a domain sentiment dictionary; S15. Quantify the sentiment corresponding to each product attribute feature word based on the domain sentiment dictionary to obtain a fine-grained sentiment value of the product attribute.

7. The system according to claim 6, wherein: The basic sentiment dictionary includes: HowNet sentiment dictionary.

8. The system according to claim 5, wherein: The nonlinear relationship between consumer satisfaction and product attribute performance measured based on the positive and negative sentiment coefficients includes: a,pos,high , P a,neg,low Two indicators are used to measure the continuity relationship between a product's attribute function and consumer satisfaction; among them, P a,pos,high =P a,pos,4 +P a,pos,5 P a,neg,low =P a,neg,3 +P a,neg,2 +P a,neg,1 Among them, P a,pos,high , P a,neg,low They represent the probability that a high score for product attribute a will result in a high total score and the probability that a low score for product attribute a will result in a low total score; P a,pos,s , P a,neg,s They represent the probability of generating a total score s when a positive evaluation and a negative evaluation are generated for attribute a, s = 1, 2, 3, 4, 5 respectively.

Images