Airport passenger service quality evaluation method based on passenger experience

Abstract

The invention discloses an airport passenger service quality evaluation method based on passenger experience. The method comprises the steps of performing preprocessing and text word segmentation processing on text information of airport passenger service quality evaluation to obtain an evaluation sample set; constructing an LDA model based on the evaluation sample set, and mining different topics and distribution probabilities thereof; determining a main topic of text evaluation according to the constructed LDA model, and calculating an emotion value, so as to obtain a passenger emotion comprehensive index of all text evaluation; obtaining an airport service guarantee comprehensive index based on a preset airport service guarantee index; and obtaining an airport passenger service quality total score by adopting a preset calculation method according to the passenger emotion comprehensive index and the airport service guarantee comprehensive index, thereby realizing airport passenger service rating of the airport. The airport passenger service quality evaluation method evaluates the airport passenger service quality from two aspects of the passenger emotion index and the service guarantee index, and the evaluation result is closer to the passenger experience and is more comprehensive.

Description

Technical Field

[0001] The present invention relates to the technical field of airport passenger service quality evaluation, and particularly relates to an airport passenger service quality evaluation method based on passenger experience. Background Art

[0002] An airport is an important place for air transportation services and a crucial link in the entire civil aviation transportation service. Its service quality largely determines the influence of the airport and also affects passengers' travel willingness, service experience, and satisfaction. In recent years, with the continuous increase in the throughput of Chinese airports and the diversified development of passenger demands, the problem of passenger complaints caused by low service quality has become prominent. Passengers' experience and evaluation are important criteria for measuring airport service quality. The most common way for airport passenger satisfaction surveys is questionnaires, but questionnaires consume high time and labor costs and are easily affected by factors such as questionnaire design, sample selection, and passengers' cooperation. With the popularity of various social media and online review websites, more and more passengers are willing and happy to post comments, psychological feelings, and negative issues on the Internet. These comments have become a more reliable source for passenger satisfaction evaluation, reflecting passengers' real psychological perception opinions without being pressured or guided by any third party. At the same time, airport passenger service quality is closely related to the airport's service guarantee ability. For example, flight delays, complaint handling, and queuing are often the main reasons for passengers' negative evaluations. Therefore, the present invention proposes to establish an airport passenger service quality evaluation method that can truly, real-time, and multi-dimensionally reflect passengers' experience feelings. Summary of the Invention

[0003] 1. Technical problems to be solved:

[0004] In view of the above technical problems, the present invention provides an airport passenger service quality evaluation method based on passenger experience, which analyzes Chinese passenger comments using general natural language processing technology, quickly obtains the emotional changes of passengers towards airport services in real time, and does not require issuing questionnaires. The obtained data is analyzed and processed, and the airport passenger service quality is evaluated from two aspects: the passenger emotion index and the service guarantee index. The evaluation results are closer to passengers' experience and more comprehensive.

[0005] 2. Technical solutions:

[0006] An airport passenger service quality evaluation method based on passenger experience, characterized by including the following steps:

[0007] Step 1: Obtain the text information on the online platform of the airport to be evaluated regarding the evaluation of airport passenger service quality, and perform preprocessing and text tokenization on the obtained text information to obtain an evaluation sample set composed of multiple text evaluations;

[0008] Step 2: Construct an LDA topic model based on the evaluation sample set; mine topics from the evaluation sample set and then summarize different topics related to the evaluation sample set and the distribution probabilities of different topics; each topic is described by its corresponding list of feature words.

[0009] Step 3: According to the constructed LDA topic model, determine the main topic of each text evaluation, calculate the sentiment value of each text evaluation, and then obtain the comprehensive passenger sentiment index of all text evaluations.

[0010] Step 4: Preset the service guarantee index of the airport, and evaluate each service guarantee index of the airport to be evaluated; comprehensively evaluate all service guarantee indexes to obtain the comprehensive index of the airport's service guarantee.

[0011] Step 5: Use a preset calculation method to obtain the total score of the airport passenger service quality by combining the comprehensive passenger sentiment index obtained in Step 3 and the comprehensive index of the airport's service guarantee obtained in Step 4, and then realize the rating of the airport passenger service of this airport.

[0012] Further, Step 1 specifically includes:

[0013] S11: Use web crawler technology to obtain the comment data of the airport to be evaluated on the online platform; the comment data includes ID number, user nickname, comment time, score, number of likes, and comment content.

[0014] S12: Clean the obtained comment data to obtain an evaluation sample set; data cleaning includes removing non-Chinese characters, removing stop words, synonym conversion, and text segmentation processing; the text segmentation processing is to perform text segmentation using the precise mode of the Jieba library in Python.

[0015] Further, Step 2 specifically includes:

[0016] S21: Perform vectorization processing on the evaluation sample set, as shown in the following formula; where x n represents the nth word segment, f n represents the word frequency corresponding to the nth word segment, and the evaluation sample set vector obtained by processing is represented by X.

[0017] X = {(x 1 , f 1 ), (x 2 , f 2 )...(x n , f n )}

[0018] S22: Perform topic modeling on the evaluation sample set vector X obtained from the analysis in step S21 using the LdaModel function in the Gensim library of Python; preset the number of topics k based on a preset method, where k is an integer; obtain the corresponding k topics Ci based on the evaluation sample set and their corresponding distribution probabilities where i represents the ordinal number corresponding to the topic and i ≤ k;

[0019] S23: Represent the list of feature words corresponding to the topic Ci in the LDA topic model result as {A i1 , A i2 , …, A ij}, where A in the list represents a feature word, its subscript i ≤ k represents the ordinal number corresponding to the topic containing the feature word; the subscript j is an integer greater than or equal to 1, representing the sorting of the frequency of occurrence of the feature word under topic i; this sorting is from large to small according to the frequency of occurrence; based on the list of feature words, the features and feature importance included in the topic Ci can be described.

[0020] Furthermore, step three specifically includes:

[0021] S31: Based on the LDA model constructed in step two, obtain all topic types corresponding to each text evaluation in the topic Ci and the distribution probability corresponding to each topic type, and take the topic type corresponding to the maximum distribution probability value as the main topic corresponding to the comment;

[0022] S32: Calculate the positive sentiment value and negative sentiment value of each text evaluation, as shown in the following formula:

[0023]

[0024] In the above formula, B pos represents the sample word library of positive emotions; B neg is the sample word library of negative emotions, and both sample databases are preset databases; P(B pos ) is the prior probability of positive emotions; P(B neg ) is the prior probability of negative emotions; x n represents the nth word segment in the word segmentation included in the comment; P(x 1 , x 2 , …, x n |B pos ) represents the probability that all the word segments included in the comment appear in the sample word library of positive emotions; P(x 1 , x 2 , …, x n |B neg ) represents the probability that the evaluation word segments appear in the negative emotion sample; P(B pos |x 1,x 2 ,…,x n ) represents the probability that the text is evaluated as a positive emotion, i.e., the positive sentiment value; P(B neg |x 1 ,x 2 ,…,x n ) represents the probability that the text is evaluated as a negative emotion, i.e., the negative sentiment value;

[0025] S33: Determine the sentiment tendency of the corresponding text evaluation by comparing the positive sentiment value and the negative sentiment value; specifically:

[0026] If P(B pos |x 1 ,x 2 ,…,x n ) > P(B neg |x 1 ,x 2 ,…,x n ), then the sentiment value Sentiment of the corresponding text evaluation = P(B pos |x 1 ,x 2 ,…,x n );

[0027] If P(B pos |x 1 ,x 2 ,…,x n ) < P(B neg |x 1 ,x 2 ,…,x n ), then the corresponding text evaluation sentiment value Sentiment = 1 - P(B neg |x 1 ,x 2 ,…,x n );

[0028] The sentiment tendency Sentiment obtained from the above comparison ∈ [0, 1]. The closer its value is to 0, the more negative the sentiment, and the closer it is to 1, the more positive the sentiment;

[0029] S34: For the sentiment tendency Sentiment of the nth text obtained in steps S32 - 33 n , convert the sentiment values of all text evaluations into sentiment index E as follows n :

[0030]

[0031] In the above formula, Sentiment n represents the sentiment tendency of the nth review text; E nRepresents the sentiment index of the nth text evaluation;

[0032] S35: Calculate the comprehensive passenger sentiment index of the airport to be evaluated:

[0033]

[0034] In the above formula, is the average value of the sentiment indices of text evaluations with the main theme C obtained for this text evaluation based on step S31; i for text evaluations with the main theme C; is the distribution probability of theme C obtained in step S22; i for theme C.

[0035] Furthermore, the service guarantee index of the preset airport in step four includes the ten-thousandths rate of complaint acceptance, flight punctuality rate, average queuing and handling time of check-in procedures during peak hours, average queuing and handling time of security checks during peak hours, and utilization rate of the waiting area during peak hours: The specific calculation of each index is as follows:

[0036] S41: Calculate the ten-thousandths rate of complaint acceptance Z as follows 1 :

[0037] The ten-thousandths rate of complaint acceptance Z 1 = Number of complaints accepted (pieces) / Passenger volume (person-times) × 10000

[0038] S42: Calculate the flight punctuality rate Z as follows 2 :

[0039]

[0040] In the above formula, a flight for which the normal flight segments meet one of the following conditions is determined as a normal flight: (1) The wheel chocks are removed within 5 minutes before and after the scheduled departure time, and the flight takes off in the positive process of flight operation without special situations such as taxiing back, aborted takeoff, return flight, and alternate landing; (2) The wheel chocks are placed no later than the scheduled arrival time;

[0041] S43: Calculate the average queuing and handling time of check-in procedures during peak hours Z as follows 3 :

[0042] The average queuing and handling time of check-in procedures during peak hours

[0043]

[0044] S44: Calculate the average queuing and handling time of security checks during peak hours Z as follows 4 :

[0045] The average queuing and handling time of security checks during peak hours

[0046]

[0047] S45: Calculate the utilization rate Z of the peak-hour machine area as follows 5 :

[0048]

[0049] In the above formula, the peak hour refers to the continuous one hour with the largest number of departing passengers arriving at the airport in a day.

[0050] Furthermore, in step four, comprehensively evaluate all service guarantee indexes to obtain the comprehensive index of airport service guarantee, which specifically includes the following steps:

[0051] S47: Convert each obtained service guarantee index from a "minimal" index to a "maximal" index as follows:

[0052] A i = max - Z i (i = 1, 3, 4, 5)

[0053] In the above formula, max represents the constraint value, and A i represents the consistency processing result corresponding to the guarantee index Z i ;

[0054] S48: Standardize each service index after consistency processing as follows to map it to the interval [0, 100]:

[0055]

[0056] In the above formula, B i represents the standardized processing result corresponding to the guarantee service index Z i ;

[0057] S49: Calculate the comprehensive index Score of airport service guarantee as follows operate :

[0058]

[0059] In the above formula, D i represents the preset weight corresponding to the guarantee service index Z i , D 1 + D 2 + D 3 + D 4 + D 5 = 1.

[0060] Furthermore, calculate the total score of airport passenger service quality in step five as follows:

[0061] Score = Score passenger ×α 1 + Score operate ×α 2

[0062] In the above formula, α 1 and α 2 are respectively the preset weight coefficients corresponding to Score passenger and Score operate , and α 1 + α 2 = 1, and the weight coefficients are determined by the analytic hierarchy process;

[0063] Rating the airport passenger service according to the total score of the airport passenger service quality, including:

[0064] If 90 ≤ Score ≤ 100, the airport passenger service quality is grade A, indicating that the airport passenger service quality is very high;

[0065] If 80 ≤ Score < 90, the airport passenger service quality is grade B, indicating that the airport passenger service quality is relatively high;

[0066] If 70 ≤ Score < 80, the airport passenger service quality is grade C, indicating that the airport passenger service quality is average;

[0067] If 60 ≤ Score < 70, the airport passenger service quality is grade D, indicating that the airport passenger service quality is poor;

[0068] If Score < 60, the airport passenger service quality is grade E, indicating that the airport passenger service quality is very poor.

[0069] Furthermore, in step S22, the preset method in the preset parameter topic number k based on the preset method is to preset manually according to experience or obtain it by calculating the consistency index; when calculating the consistency index, the one with the highest consistency is selected as the optimal topic number k.

[0070] 3. Beneficial effects:

[0071] (1) The method for evaluating the quality of airport passenger service based on passenger experience provided by the present invention uses the LDA topic model analysis method to mine and process the evaluation samples to obtain all the topics related to the comments of the airport to be evaluated. The LDA model is based on Bayesian statistical theory, regards each group of documents as a multinomial distribution of different topics, and each topic as a multinomial distribution of different words, and obtains the association between the document and the topic through the statistical inference result, and its analysis result quality is relatively high.

[0072] (2) In a method for evaluating the quality of airport passenger services based on passenger experience in the present invention, the sentiment tendency of each passenger comment is calculated, and the comprehensive passenger sentiment index is obtained after synthesis; this process can relatively accurately and quickly obtain the sentiment of passengers towards airport services.

[0073] (3) In the present invention, the airport is evaluated from multiple dimensions such as airport transportation, check-in procedures, delay services, catering and shopping, environmental facilities, etc., which helps airport managers identify specific aspects that need to be improved and improve and enhance the quality of airport passenger services. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 is the overall flowchart of this method;

[0075] Figure 2 is the schematic diagram of the composition of the total score of airport passenger service quality in this solution. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] The present invention will be specifically described below with reference to the accompanying drawings.

[0077] As shown in the attached Figure 1 、 2 figures, a method for evaluating the quality of airport passenger services based on passenger experience is characterized by including the following steps:

[0078] Step 1: Obtain the text information on the evaluation of airport passenger service quality in the online platform of the airport to be evaluated, and perform preprocessing and text word segmentation on the obtained text information to obtain an evaluation sample set composed of multiple text evaluations;

[0079] Step 2: Based on the evaluation sample set, construct an LDA topic model, mine topics from the evaluation sample set, and then summarize different topics and the distribution probabilities of different topics; each topic is described by its corresponding list of feature words;

[0080] Step 3: According to the constructed LDA topic model, determine the main topic of each text evaluation, calculate the sentiment value of each text evaluation, and then obtain the comprehensive passenger sentiment index of all text evaluations;

[0081] Step 4: Preset the service guarantee index of the airport, and evaluate each service guarantee index of the airport to be evaluated; comprehensively evaluate all service guarantee indexes to obtain the comprehensive index of the airport service guarantee;

[0082] Step 5: Use a preset calculation method to obtain the total score of airport passenger service quality by combining the comprehensive passenger sentiment index obtained in Step 3 with the comprehensive index of airport service guarantee obtained in Step 4, and then realize the rating of airport passenger services of this airport. Specific embodiments:

[0084] As shown in the appendix Figure 2 This specific embodiment takes the comments on airports on the Dianping website as an example to elaborate on this method in detail.

[0085] I. Data acquisition

[0086] Comment data of the top 15 airports in terms of passenger throughput in China in 2023 was crawled from the Dianping website. There were a total of 51,883 comment data from 2015 to 2024. Each piece of data included an ID number, user nickname, comment time, score, number of likes, and comment content.

[0087] II. Data cleaning

[0088] The obtained comment data was cleaned as follows: (1) Remove non-Chinese characters. Remove English, numbers, emojis, and other special characters in the comments, and only retain Chinese characters. (2) Remove stop words. By loading a stop word dictionary, words irrelevant to airport services were removed. The stop word dictionary can be a commonly used Chinese stop word dictionary (Baidu stop dictionary, Harbin Institute of Technology stop dictionary), and on this basis, high-frequency words irrelevant to airport services were removed, such as "airport", "China", "taking a plane", "hahaha", etc. (3) Synonym conversion. To improve the correctness and accuracy of the results, a synonym conversion list related to airport services was set. For example, replace "personnel", "little sister", "flight attendant", "air stewardess", etc. with "staff", and replace "beef noodles", "cold noodles with sauce", "hot pot", "noodles", "food", etc. with "catering". (4) Text tokenization. The exact mode of the Jieba library in Python was used for text tokenization. Since the comments contain a large number of place names, airline names, etc. that are meaningless for airport service analysis, during the tokenization process, words with the part-of-speech judged as "institutional organization (nt)" and "place name (ns)" were filtered out. At the same time, single characters with a character length of 1 were filtered out to improve the accuracy of text analysis. In addition, to prevent specific words from being separated during tokenization, some special words related to airport services were added, such as "online car-hailing", "high-end and elegant", "checked luggage", etc.

[0089] After the above text cleaning steps, a total of 49,823 pieces of data remained.

[0090] Taking the comparison before and after cleaning of a certain comment as an example is as follows:

[0091] Original comment: At the only airport in Hangzhou, the airport facilities are quite good, and the cleaning of the cleaners is also very timely. The drinking water used in the airport is direct drinking water of Smith, and it makes people feel very safe to drink. Today is the second day after the festival, so there are not many people, and the whole airport looks relatively orderly. It's just that you need to take the airport bus directly for transportation, but it doesn't matter then because there is a direct airport express line.

[0092] After cleaning: The only device is good for cleaning and sanitation. Drinking water and direct drinking water are provided by Smith. It will be okay by then. The transportation bus will reach directly. It doesn't matter at that time. It will reach directly.

[0093] III. Determine the number of topics

[0094] In this embodiment, topic consistency is used as an indicator to judge the optimal number of topics in the evaluation sample set. The value of the consistency index is calculated through the Gensim library of Python. The number of topics corresponding to the highest value is the best number of topics.

[0095] If the topic clustering effect is not obvious, or most of the feature words cannot be reasonably explained, data cleaning should be performed again. Stopwords that are meaningless can be added, a list of synonym conversions can be added, and special words for airport services can be added. Until the topic clustering effect is obvious and most of the feature words are highly interpretable.

[0096] IV. Determine topics using the LDA topic model

[0097] The optimal number of topics for the LDA model is calculated through the Gensim library of Python. When the number of topics is 10, the consistency value is the highest, achieving the optimal model. Subsequently, set the LDA model parameters: the number of topics is 10, and the number of training times is 30. Finally, the results of the LDA model are obtained as follows:

[0098] Topic 1: Urban subway, bus, transportation, distance, arrival, direct line, facility, taxi Topic 2: Security check, boarding, check-in, checked baggage, queue, self-service handling, time, channel Topic 3: Hours, really, delay time, boarding in advance, evening, transit, flight ticket cancellation Topic 4: Feeling, really, like, hope, travel, every time, look forward to, mood, place, impression Topic 5: Catering, price, coffee, good, brand, delicious, time, taste, fast food, store Topic 6: Good, feeling, clean, environment, overall, facility, toilet, very big, especially complete Topic 7: Staff, service, work, service attitude, enthusiastic, experience, provide, passenger, ground crew

[0099] Patient

[0100] Topic 8: Baggage, exit, arrival, landing, pick-up, carousel, service desk, get a trolley, online car-hailing

[0101] Topic 9: Parking lot, pick-up, parking, ticket change, underground, charge, weekend, advertisement promotion

[0102] Unlicensed taxi

[0103] Topic 10: Design, trypophobia, architecture, technology, shape, best, ceiling, update, terminal

[0104] The above 10 topics are respectively described as airport transportation, check-in procedures, flight delays, overall feelings, catering and shopping, environmental facilities, staff, baggage claim, pick-up, and terminal design.

[0105] V. Emotional Index Calculation

[0106] As in steps S32 - 34, the emotional index of each comment was calculated using the SnowNLP library in Python, and the comprehensive emotional index of airport passengers was obtained by summarizing according to the main themes of each comment.

[0107] In this embodiment, the example of the main theme results obtained based on step S31 is as follows:

[0108] Example of the main theme results of each evaluation text

[0109]

[0110]

[0111] The example of the calculation results in steps S32 - S33 is as follows:

[0112] Example of the calculation results of the emotional tendency of each evaluation text

[0113]

[0114]

[0115] The example of the calculation results in step S34 is as follows:

[0116] Example of the calculation results of the emotional index of each evaluation text

[0117]

[0118]

[0119] The example of the calculation results in step S35 is as follows:

[0120] Example of the calculation of the comprehensive emotional index of passengers

[0121]

[0122]

[0123] VI. Example of Determining the Airport Guarantee Index and Calculating the Comprehensive Index of Airport Service Guarantee

[0124] In this method, based on "Civil Aviation Transportation Airport Service Quality (MH / T 5104-2013)" and "Measures for Statistics of Civil Aviation Flight Regularity", from the perspective of airport management, aspects closely related to passenger services such as flight delays, complaint handling, and queuing waiting are selected. Considering the measurability, representativeness, and comprehensiveness of the indicators, 5 airport service guarantee indexes are determined. The guarantee indexes defined in this method include the ten-thousandths rate of complaint acceptance, flight regularity rate, average queuing and handling time of boarding procedures during peak hours, average queuing and handling time of security check during peak hours, and utilization rate of the waiting area during peak hours.

[0125] According to the monthly "Circular on the Situation of Passenger Service Complaints in Public Air Transportation" issued by the Civil Aviation Administration of China, the number of complaints accepted by each airport in the current month can be obtained, and the cumulative number of complaints accepted in the current year can be obtained by accumulation. According to the "Statistical Bulletin of Civil Aviation Transportation Airport Production in the Whole Country" issued by the Civil Aviation Administration of China every year, the annual passenger transportation volume of each airport can be obtained.

[0126] Taking Guangzhou Baiyun International Airport, Shanghai Pudong International Airport, and Beijing Capital International Airport in 2022 as examples, the ten-thousandths rate of complaint acceptance Z 1 The calculation example is as follows:

[0127] Number of complaints accepted (pieces) Volume of passenger transport (person-times) <![CDATA[Complaint acceptance rate per ten thousand Z 1 > Guangzhou / Baiyun 69 26104989 0.0264 Shanghai / Pudong 69 14178386 0.0487 Beijing / Capital 7 12703342 0.0055

[0128] Taking October 2024 as an example, the planned number of flights at Beijing Capital International Airport is 38,688 flights, and the number of regular flights is counted as 35,523 flights. The flight regularity rate Z 2 is 91.81%.

[0129] The peak hour is the period with the highest outbound passenger flow at the airport. Conducting on-site observations at the airport during the peak period can obtain the time for each passenger to complete the boarding procedures, the time to start queuing, and the total number of people queuing for boarding procedures during peak hours. Taking Terminal 2 of Beijing Capital International Airport as an example, its daily peak outbound passenger periods are generally from 8:00 to 10:00 in the morning and from 14:00 to 15:00 in the afternoon. A certain check-in counter at Terminal 2 of Beijing Capital International Airport is investigated during these two periods respectively, and the investigation form is as follows:

[0130] Passenger Time to start queuing Time to complete check-in 1 7:55 8：06 2 7:56 8：10 3 7:56 8：10 4 7:57 8:10 5 7:58 8:11 …… …… …… 221 8:58 9:17

[0131] In this embodiment, the average queuing and handling time of boarding procedures during peak hours at Beijing Capital International Airport

[0132] The average queuing and handling time of security check during peak hours Z 4Reflects the efficiency and service level of airport security check services during peak hours. Conducting on-site observations of the airport during peak hours can obtain the time taken for each passenger to complete the security check, the time when the security check queue starts, and the total number of people queuing for the security check during peak hours. Taking the T2 terminal of Beijing Capital International Airport as an example, the peak departure passenger hours are generally from 8:00 to 10:00 in the morning and from 2:00 to 3:00 in the afternoon. Investigations were conducted on a certain security check channel at the T2 terminal of Beijing Capital International Airport during these two time periods respectively. The investigation form is as follows:

[0133]

[0134]

[0135] In this embodiment, Beijing Capital International Airport

[0136] Average queuing and processing time for security check during peak hours

[0137] Conducting on-site observations of the airport waiting area during peak hours can obtain the total number of waiting passengers during peak hours. Taking the waiting area at boarding gates 19 - 22 of the T2 terminal of Beijing Capital International Airport as an example, the peak departure passenger hours are generally from 8:00 to 10:00 in the morning and from 2:00 to 3:00 in the afternoon. Investigations were conducted on the waiting area at boarding gates 19 - 22 of the T2 terminal of Beijing Capital International Airport during these two time periods respectively. There are 132 seats in this area, 144 waiting passengers at 8:00 in the morning, and the utilization rate Z of the waiting area during peak hours 5 is 1.09.

[0138] Calculation example of the consistency and standardization processing of indicators Z1 - Z5 at Beijing Capital International Airport:

[0139]

[0140] The result of standardization processing is:

[0141]

[0142]

[0143] VII. Calculating the total score of airport passenger service quality

[0144] Taking Beijing Capital International Airport as an example, based on the previously calculated passenger emotion comprehensive index of 85.708 and service guarantee comprehensive index of 88.44, determining the weight coefficients a1 as 0.6 and a2 as 0.4, the total score of airport passenger service quality can be calculated as 85.708 * 0.6 + 88.44 * 0.4 = 86.8. The airport passenger service quality is at level B, indicating that the airport passenger service quality is relatively high.

[0145] Although the present invention has been disclosed above in preferred embodiments, they are not intended to limit the present invention. Any person skilled in the art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be defined by the protection scope of the claims of this application.

Claims

1. A method for evaluating airport passenger service quality based on passenger experience, characterized in that: The following steps are involved: Step 1: Obtain text information about the airport passenger service quality evaluation on the online platform of the airport to be evaluated, and perform preprocessing and text segmentation on the obtained text information to obtain an evaluation sample set consisting of multiple text evaluations; Step 2: Construct an LDA topic model based on the evaluation sample set; mine topics from the evaluation sample set and summarize different topics related to the evaluation sample set and the distribution probability of different topics; each topic is described by its corresponding feature word list; Step 3: According to the constructed LDA topic model, determine the main topic of each text evaluation, and calculate the sentiment value of each text evaluation, and then obtain the comprehensive passenger sentiment index of all text evaluations; Step 4: preset the service guarantee index of the airport, and evaluate each service guarantee index of the airport to be evaluated; comprehensively evaluate all service guarantee indexes to obtain the comprehensive index of the airport's service guarantee; Step 5: The passenger sentiment comprehensive index obtained in step 3 and the airport service guarantee comprehensive index obtained in step 4 are calculated using a preset method to obtain the total score of the airport passenger service quality, thereby achieving the airport passenger service rating for the airport.

2. The method for evaluating airport passenger service quality based on passenger experience according to claim 1, characterized in that: Step 1 specifically includes: S11: Using crawler technology to obtain review data of the airport to be reviewed on the online platform; the review data includes ID number, user nickname, review time, score, likes and review content; S12: Perform data cleaning on the acquired comment data to obtain an evaluation sample set; data cleaning includes removing non-Chinese characters, removing stop words, synonym conversion and text segmentation processing; the text segmentation processing is to perform text segmentation using the precise mode of Python's Jieba library.

3. The method for evaluating airport passenger service quality based on passenger experience according to claim 1, characterized in that: Step 2 specifically includes: S21: Vectorize the evaluation sample set as shown in the following formula; where x n Indicates the nth participle, f n represents the word frequency corresponding to the nth word segmentation, and the evaluation sample set vector obtained by processing is represented by X; X={(x1,f1),(x2,f2)…(x n ,f n )} S22: Perform topic modeling on the evaluation sample set vector X obtained by the analysis in step S21 through the LdaModel function in Python's Gensim library; preset the number of topic parameters k based on the preset method, where k is an integer; obtain the corresponding k topics Ci based on the evaluation sample set and their corresponding distribution probabilities Where i represents the ordinal number corresponding to the topic and i≤k; S23: The feature word list corresponding to the topic Ci in the LDA topic model result is replaced by {A i1 , A i2 , …, A ij } indicates that A in the list represents a feature word, and its subscript i≤k represents the ordinal number corresponding to the topic containing the feature word; the subscript j is an integer greater than or equal to 1, which represents the order of the frequency of occurrence of the feature word under topic i; the order is from high to low frequency; based on the feature word list, it is possible to describe the features included in the topic Ci and the importance of the features.

4. The method for evaluating airport passenger service quality based on passenger experience according to claim 3 is characterized by: Step three specifically includes: S31: Based on the LDA model constructed in step 2, all topic types corresponding to the topic Ci of each text evaluation and the distribution probability corresponding to each topic type are obtained, and the topic type corresponding to the maximum distribution probability value is taken as the main topic corresponding to the comment; S32: Calculate the positive sentiment value and negative sentiment value of each text evaluation, as follows: In the above formula, B pos Sample word library representing positive emotions; B neg is a sample word library of negative emotions. Both sample databases are preset databases. pos ) is the prior probability of positive emotion; P(B neg ) is the prior probability of negative emotions; x n represents the nth word in the comment; P(x1,x2,…,x n |B pos ) represents the probability of all the words contained in the comment appearing in the sample vocabulary of positive emotions; P(x1,x2,…,x n |B neg ) represents the probability of the evaluation word appearing in the negative emotion sample; P(B pos |x1,x2,…,x n ) represents the probability that the text is evaluated as positive, that is, the positive sentiment value; P(B neg |x1,x2,…,x n ) represents the probability that the text is evaluated as negative, i.e., the negative sentiment value; S33: Determine the sentiment tendency of the corresponding text evaluation by comparing the positive sentiment value and the negative sentiment value; specifically: If P(B pos |x1,x2,…,x n )>P(B neg |x1,x2,…,x n ), then the corresponding sentiment value of the text evaluation is Sentiment=P(B pos |x1,x2,…,x n ); If P(B pos |x1,x2,…,x n ) <P(B neg |x1,x2,…,x n ), then the corresponding text evaluation sentiment value Sentiment=1-P(B neg |x1,x2,…,x n ); The sentiment tendency obtained by the above comparison is Sentiment∈[0,1]. The closer the value is to 0, the more negative the sentiment is, and the closer it is to 1, the more positive the sentiment is. S34: Get the sentiment of the nth text in steps S32-33 n , the sentiment values ​​of all text reviews are converted into sentiment index E as follows n : In the above formula, Sentiment n Indicates the sentiment tendency of the nth comment text; E n Represents the sentiment index of the nth text evaluation; S35: Calculate the comprehensive passenger sentiment index of the airport to be evaluated: In the above formula, The main theme obtained in step S31 is evaluated as C for the text. i The mean sentiment index of text evaluation; is the subject C obtained in step S22 i The distribution probability of .

5. The method for evaluating airport passenger service quality based on passenger experience according to claim 4, characterized in that: The service guarantee index of the preset airport in step 4 includes the complaint acceptance rate, flight regularity rate, average queuing and processing time for check-in during peak hours, average queuing and processing time for security check during peak hours, and airport area utilization rate during peak hours. The specific calculation of each index is: S41: Calculate the complaint acceptance rate Z1 as follows: Complaint acceptance rate Z1 = number of complaints accepted (cases) / passenger transportation volume (person-times) × 10,000 S42: Calculate the flight regularity rate Z2 as follows: In the above formula, a flight that meets one of the following conditions is considered a normal flight: (1) the flight is unblocked within 5 minutes before or after the scheduled departure time and takes off according to the forward flight operation process without any special circumstances such as taxiing back, aborted takeoff, return flight, or alternate landing; (2) the flight is unblocked no later than the scheduled arrival time; S43: The average queuing and processing time Z3 for check-in during peak hours is calculated as follows: S44: The average queuing and processing time Z4 for security check during peak hours is calculated as follows: S45: Calculate the machine area utilization rate Z5 during the peak hour as follows: In the above formula, peak hour refers to the continuous hour in a day when the number of departing passengers arriving at the airport is the largest.

6. The method for evaluating airport passenger service quality based on passenger experience according to claim 5, characterized in that: In step 4, all service guarantee indexes are comprehensively evaluated to obtain the comprehensive index of airport service guarantee, which specifically includes the following steps: S47: Convert each obtained service guarantee index from a "minimum" index to a "maximum" index as follows: A i =max-Z i (i=1,3,4,5) In the above formula, max represents the constraint value, A i Indicates the security index Z i The corresponding consistency processing results; S48: The service indexes that have undergone consistency processing are standardized as follows, so as to be mapped to the interval [0,100]: In the above formula, B i Indicates the guarantee service index Z i The corresponding standardized processing results; S49: The comprehensive index score of the field service guarantee is calculated as follows: operate : In the above formula, D i Represents the preset guarantee service index Z i The corresponding weight is D1+D2+D3+D4+D5=1.

7. The method for evaluating airport passenger service quality based on passenger experience according to claim 6, characterized in that: In step 5, the total score of airport passenger service quality is calculated as follows: Score=Score passenger ×α1+Score operate ×α2 In the above formula, α1 and α2 are the preset scores. passenger and Score operate The corresponding weight coefficient is α1+α2=1, and the weight coefficient is determined by the hierarchical analysis method; Rating airport passenger services based on the overall score of the airport passenger service quality, including: If 90≤Score≤100, the airport passenger service quality is grade A, which means that the airport passenger service quality is very high; If 80≤Score<90, the airport passenger service quality is B, indicating that the airport passenger service quality is relatively high; If 70≤Score<80, the airport passenger service quality is C, which means the airport passenger service quality is average; If 60≤Score<70, the airport passenger service quality is D, indicating that the airport passenger service quality is poor; If Score<60, the airport passenger service quality is E, which means that the airport passenger service quality is very poor.

8. The method for evaluating airport passenger service quality based on passenger experience according to claim 3, characterized in that: In step S22, the preset method in the preset parameter topic number k based on the preset method is manually preset based on experience or obtained by calculating the consistency index; When calculating the consistency index, the one with the highest consistency is selected as the optimal number of topics k.

Images