Personalized travel route planning method considering multiple themes

Through the LDA model and TF-IDF algorithm combined with entropy weight method to evaluate user preferences and scenic spot characteristics, combined with the genetic algorithm optimization model, the problems of insufficient user data and difficult to meet multi-theme needs in the existing technology are solved, and more accurate and personalized The travel route planning improves user satisfaction and evaluation accuracy.

CN118981567AActive Publication Date: 2024-11-19NANTONG UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411004094.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-19
Estimated Expiration
2044-07-25

AI Technical Summary

Technical Problem

Existing personalized travel route planning methods are difficult to accurately identify user preferences with less user data and are difficult to consider multiple topic needs at the same time, often overlooking or limited to planning for a single topic.

Method used

The LDA model and TF-IDF algorithm are used to calculate user preferences and scenic spot theme characteristics, and the entropy weight method is used to evaluate user satisfaction. The model is optimized through two-stage method and genetic algorithm to meet the multi-theme needs given by users and plan to include multiple categories. Travel route.

Benefits of technology

It improves the degree of personalization of tourism planning when the individual attributes of users are missing, and improves the accuracy of attraction assessment and user satisfaction through multi-dimensional evaluation, avoids information overload, and provides a more intuitive and practical travel route.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981567A_ABST
    Figure CN118981567A_ABST
Patent Text Reader

Abstract

The invention provides a personalized travel route planning method considering multiple themes, and belongs to the technical field of traffic planning design. The technical problem that the travel theme is not considered in the travel route or the travel theme is single is solved. According to the technical scheme, the method comprises the following steps: S1, obtaining user travel data and scenic spot data; s2, calculating user travel preferences, and dividing scenic spot themes; s3, calculating the score of each user on each scenic spot; s4, establishing a multi-topic limitation optimization model; and S5, solving the model by using a two-stage method and a genetic algorithm. The method has the advantages that on the basis of considering user preferences, theme requirements are further met, limiting conditions such as time are set, suggestions of personalized travel routes are provided for users, and therefore the satisfaction degree of the users to the travel routes is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of traffic planning and design, and in particular to a personalized travel route planning method considering multiple topics. Background Art

[0002] Themed tourism is very popular, which shows that the formulation of personalized travel routes is no longer limited to traditional scenic spots, but pays more attention to the personal interests and needs of tourists. Therefore, how to formulate multi-themed personalized travel routes has become a hot topic at present.

[0003] The research on personalized travel route planning mainly focuses on how to accurately obtain user preferences and meet single-theme or no-theme requirements. Design and Implementation of a Personalized Tourism Recommendation System Based on the Data Mining and Collaborative Filtering Algorithm starts from the perspective of mining similarities between users, calculates user similarities, and recommends attractions visited by people with high similarity; A Picture-Based Approach to Tourism Recommendation System uses an image-based questionnaire survey method as an implicit induction of user preferences for tourism products to accurately recommend products to travelers with different tastes; A Comprehensive Survey on Travel Recommender Systems recommends restaurants, museums, and hotels based on user preferences. At present, the main personalized travel route planning relies on a large number of users to improve the accuracy of identifying user preferences, but it is difficult to cope with the situation of less user data; and in terms of considering theme requirements, either ignore theme requirements, or only consider specific categories of planning under a single theme, such as only recommending restaurants. Summary of the invention

[0004] The purpose of the present invention is to provide a personalized travel route planning method considering multiple topics. Based on comment text data, the LDA model and TF-IDF algorithm are used to calculate user preferences and thematic characteristics of scenic spots. In the user's satisfaction with the scenic spot, not only the matching degree of user preferences and scenic spot characteristics is considered, but also the entropy weight method is used to calculate the fixed scores under indicators such as the number of repeat customers, local people's ratings, and the number of people who check in in the past 30 days; in the optimization model, with the goal of maximizing user satisfaction, a constraint condition on the number of topics is established to meet the user's given topic requirements, and a two-stage method is used to simplify the model; on the basis of comprehensive scenic spot characteristics and user preferences, the present invention further meets the topic requirements and plans travel routes containing multiple categories.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is specifically: a personalized travel route planning method considering multiple topics, comprising the following steps:

[0006] Step S1, obtaining user travel data and scenic spot data;

[0007] Step S2, calculating the user's travel preferences and dividing the scenic spots into themes;

[0008] Step S3, calculating the ratings of each user for each scenic spot;

[0009] Step S4, establishing a multi-topic restriction optimization model;

[0010] Step S5, solving the model using a two-stage method and a genetic algorithm.

[0011] Preferably, in step S1, the specific process of obtaining the user's travel and scenic spot data is:

[0012] Step S11, obtaining scenic spot data such as comprehensive scores, evaluation data sets PD, recommended products, and regions through channels such as Meituan, Dianping, and Douyin, to form a scenic spot set A;

[0013] Step S12, obtaining user travel data such as the region, user historical rating data R and comment text data UD through Meituan, Dianping, Douyin and other channels;

[0014] Specifically, in step S2, the specific process of calculating the user's travel preferences and dividing the scenic spot themes is as follows:

[0015] Step S21, use the TF-IDF algorithm to process the UD and calculate the user preference. After the algorithm processing, s high-frequency keywords are extracted from the UD to form the user preference attribute set I= <I1,I2,……,I s >, its corresponding preference value is i= <i1,i2,……,i s >, where I s is the word to be calculated, j is the comment containing the word, j∈UD, For the I in comment j s The number of is the total number of words in comment j.

[0016] Step S22, based on the scenic spot evaluation data set PD, use the LDALDA topic model to extract the sub-topics and their keywords of each scenic spot respectively, analyze the keyword extraction results of each scenic spot one by one, remove irrelevant words, modify the number of topics one by one, but do not modify the number of keywords, so as to be more in line with the actual sub-topics.

[0017] Step S23, clustering the sub-topics into n categories according to the subject words by using the K-means clustering method, and calculating the relevance to each topic, the specific process is as follows:

[0018] Step S231, randomly selecting the initial centers of n classes;

[0019] Step S232: In the i-th iteration, for any sample, calculate its Euclidean distance to the n centers according to the word frequency, and classify the sample into the class where the center with the smallest distance is located;

[0020] Step S233, using the mean method or other methods to update the center value of the class;

[0021] Step S234, looping steps S232 and S233, if the center values ​​of the n cluster centers remain unchanged, the iteration ends; Step S235, comprehensively adjusting and determining the topic label according to the actual situation, and further calculating the relevance of each keyword of each scenic spot in its evaluation data set according to the TD-IDF algorithm, the topic label is K= <K1,K2,……,K n >, the corresponding correlation is k = <k1,k2,……,k n >, where k n Keyword K n 's relevance.

[0022] Step S3, the specific process of calculating the score of each scenic spot is as follows:

[0023] Step S31, calculate the relationship between user u and scenic spot A i The interest matching score is:

[0024] Step S32: Combine the local ratings, number of repeat visitors, number of check-ins in the past 30 days, and other indicators of the scenic spot, and use the entropy weight method to calculate the fixed rating of the i-th scenic spot. The calculation formula is r i =∑ j w j ·a ij , where w j =(1+(lnn) -1 ·∑ i p ij ·lnp ij )·(∑ j w j b ij ) -1 , a ij is the value of the jth indicator of the ith scenic spot, b ij for a ijNormalized value; Specifically, step S4, the specific process of establishing the multi-topic restriction optimization model is as follows; step S41, considering the actual travel situation of the user, the following assumptions are made for the model:

[0025] Step S411: The starting point of the user's travel route is A1, the set of candidate attractions is A, and the number of candidate attractions is N. A .

[0026] Step S412: If the evaluation given by each user to each scenic spot is true and reliable and reflects his or her true preferences, then user u's evaluation of scenic spot A is i Rating where r i u For user u to visit attraction A i Interest matching score;

[0027] Step S42, with the goal of maximizing user satisfaction, the objective function can be established as where y i ={0,1} indicates whether attraction A is selected in the route i ;

[0028] Step S43, considering the actual travel situation of the user, the following restrictions are imposed on the model:

[0029] Step S431: The longest playing time of the user for this trip is T max , since the traffic time is random, chance constrained programming is adopted, and the time window constraint is in v i For Attraction A i The residence time, t ij For Attraction A i With attractions A j The travel time between

[0030] Step S432: for the desired topic label B given by the user, the route contains at most S max attractions that do not meet the theme requirements, then the theme constraint is in B≥1, if attraction A i If the topic tag meets the requirements, K i For Attraction A i Collection of hashtags

[0031] Step S433: The user's travel route must form a loop, so a constraint is constructed.

[0032] Step S434: Each scenic spot can only be visited once, so constraints (1-3) are constructed:

[0033]

[0034] μ i -μ j +1≤(N A -1)(1-x ij ), (2)

[0035] 2≤μ i ≤N A , (3)

[0036] Where k = 2, ..., N A ,μ i A i The order of visits in the route, i=2,...,N A ;

[0037] Step S435: Only the selected scenic spots can form a route, so the constraints can be constructed. Where i=1,……,N A ;

[0038] Specifically, in step S5, the specific process of solving the model using the two-stage method is as follows:

[0039] Step S51, establishing the first stage model, the specific process of obtaining the optimal scenic spot visit set is as follows:

[0040] Step S511, the objective function is simplified to

[0041] Step S512, relax the constraints. Step S431 can be transformed into The constraint condition of step S432 can be transformed into Among them, M is an infinite number and Z is an auxiliary variable.

[0042] Step S513, using a genetic algorithm to solve the first stage model;

[0043] Step S52, establish the second stage model, and determine the specific process of the access sequence as follows:

[0044] Step S521, the objective function is simplified to

[0045] Step S522, strengthen the constraint conditions. Step S431 can be converted into Step S432 can be converted into

[0046] Step S523, the genetic algorithm solves the second stage model.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The present invention uses the LDA model and TF-IDF algorithm to accurately calculate the user's preferences and the theme characteristics of each scenic spot through comment text data. Traditional methods usually only rely on users' simple ratings or recommendation systems, while the present invention makes users' needs more accurately matched with actual tourism resources by deeply analyzing the theme information hidden in the text data. This method can improve the personalization of tourism planning in the absence of individual user attributes.

[0049] 2. When evaluating user satisfaction with scenic spots, the present invention introduces multi-dimensional considerations, such as the number of repeat customers, local ratings, and the number of people who checked in in the past 30 days. Compared with the traditional single-indicator evaluation method, this comprehensive evaluation model can more comprehensively reflect the attractiveness and actual value of scenic spots. The fixed score is calculated by the entropy weight method, which further improves the objectivity and accuracy of the evaluation, thereby providing users with more reliable travel suggestions.

[0050] 3. In the process of optimizing the tourist route, the present invention sets the constraint of the number of themes with the goal of maximizing user satisfaction, and adopts a two-stage method to simplify the solution model on this basis. This optimization strategy not only takes into account the theme requirements given by the user, but also effectively avoids the problems of information overload or insufficient recommendations that may occur in tourist planning through the introduction of constraints; by simplifying the model, the final tourist route planning is more intuitive and practical, greatly improving the user experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0052] Figure 1 The present invention is a flowchart of a method for personalized travel route planning that takes multiple topics into consideration, which is proposed in an embodiment of the present invention.

[0053] Figure 2 This is a flow chart of calculating user travel preferences and dividing scenic spot keywords in the method proposed in an embodiment of the present invention.

[0054] Figure 3 This is a flow chart of calculating the score of each user for each scenic spot in the method proposed in an embodiment of the present invention.

[0055] Figure 4 This is a flow chart of using a two-stage method and a genetic algorithm to solve a model in the method proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0057] Example 1

[0058] See also Figures 1 to 4 The technical solution provided by this embodiment is a personalized travel route planning method considering multiple topics, comprising the following steps:

[0059] Step S1, obtaining user travel data and scenic spot data;

[0060] Step S2, calculating the user's travel preferences and dividing the scenic spots into themes;

[0061] Step S3, calculating the ratings of each user for each scenic spot;

[0062] Step S4, establishing a multi-topic restriction optimization model;

[0063] Step S5, solving the model using a two-stage method and a genetic algorithm.

[0064] Preferably, in step S1, the specific process of obtaining the user's travel and scenic spot data is:

[0065] Step S11, obtain the scenic spot data of Xianju City through the Dianping website. Among them, the types of scenic spots include restaurants, shopping malls, surrounding tours, etc. In this embodiment, restaurants are mainly used as the main scenic spots, forming set A. Its data includes comprehensive scores, evaluation data sets PD, recommended products, etc.;

[0066] Step S12, obtaining travel data of Dianping users, such as the region, user historical rating data R and comment text data UD, etc., by crawling;

[0067] Specifically, in step S2, the specific process of calculating the user's travel preferences and dividing the scenic spot themes is as follows:

[0068] Step S21, use the jieba word segmentation package in Python to process the user comment text data UD, and use the TF-IDF algorithm to calculate the word frequency. After algorithm processing, s high-frequency keywords are extracted from the comment text data UD to form the user preference attribute set I= <I1,I2,……,I s >, its corresponding preference value is i= <i1,i2,……,i s >. Among them, I s is the word to be calculated, j is the comment containing the word, j∈UD, For the I in comment js The number of is the total number of words in comment j.

[0069] Step S22, taking restaurants as an example, use jieba to perform word segmentation on the restaurant evaluation dataset PD, and use the LDA topic model to extract the sub-topics and key words of each restaurant. Analyze the extraction results one by one, remove irrelevant words, modify the number of topics one by one, but do not modify the number of key words.

[0070] Step S23, clustering the sub-topics into four categories according to the subject words by using the K-means clustering method, and calculating the relevance to each topic. The specific process is as follows:

[0071] Step S231, randomly select the initial centers of 4 classes;

[0072] Step S232: In the i-th iteration, for any sample, calculate its Euclidean distance to the four centers respectively according to the word frequency, and classify the sample into the class where the center with the smallest distance is located;

[0073] Step S233, using the mean method or other methods to update the center value of the class;

[0074] Step S234, looping steps S232 and S233, if the center values ​​of the four cluster centers remain unchanged, the iteration ends;

[0075] Step S235, according to the actual situation, comprehensively adjust and determine the theme tags, which are "environment", "service", "taste", "average price", "party", etc., and further calculate the relevance of each theme word of each scenic spot in its evaluation data set according to the TF-IDF algorithm, then the theme tag is K= <K1,K2,……,K n >, the corresponding correlation is k = <k1,k2,……,k n >, where k n Keyword K n 's relevance.

[0076] Step S3, the specific process of calculating the rating of each restaurant is as follows;

[0077] Step S31, calculate the interest matching score between user u and restaurant A1, the formula is

[0078] Step S32: Combine the three indicators of the restaurant's local rating, number of repeat customers, and number of people who checked in in the past 30 days, and use the entropy weight method to calculate the restaurant's local rating.

[0079] Calculate the fixed score of the i-th restaurant, the calculation formula is r i =∑ j w j ·a ij, where w j =(1+(lnn) -1 ·∑ i p ij ·lnp ij )·(∑ j w j b ij ) -1 , a ij is the value of the jth indicator of the i-th restaurant, b ij for a ij Normalized value, k = (lnn) -1 .

[0080] Specifically, step S4, the specific process of establishing the multi-topic restriction optimization model is as follows;

[0081] Step S41, considering the actual travel situation of the user, the following assumptions are made for the model:

[0082] Step S411: The starting point of the user's travel route is A1, the set of candidate restaurants is A, and the number of candidate attractions is N. A .

[0083] Step S412: The evaluations given by each user to each restaurant are authentic and reliable, and reflect their true preferences. Then user u's evaluation of restaurant A i Rating where r i u For user u to restaurant A i Interest matching score;

[0084] Step S413: The starting point of the user's travel route is A1, the candidate scenic spot set is A, the number of restaurants is 200, N A =201.

[0085] Step S42, with the goal of maximizing user satisfaction, the objective function can be established as where y i ={0,1} indicates whether restaurant A is selected in the route i

[0086] Step S43, considering the actual travel situation of the user, the following restrictions are imposed on the model:

[0087] Step S431: The longest playing time of the user for this trip is T max = 12h. Since the traffic time is random, chance-constrained planning is adopted, and the time window constraint is in v i For Restaurant A iRestaurant A i With Restaurant A j The longitude and latitude of the vehicle are used as input, and the distance is obtained by calling the Amap API. Based on the average vehicle speed, the travel time t can be calculated. ij ;

[0088] Step S432: for the desired theme tag B given by the user, which is {"team building", "party"}, the route contains at most 2 restaurants that do not meet the theme requirements, then the theme constraint is in B≥1, if restaurant A i If the topic tag meets the requirements, K i For Restaurant A i Collection of hashtags

[0089] Step S433: The user's travel route must form a loop, so a constraint is constructed.

[0090] Step S434: Each scenic spot can only be visited once, so the following constraints (1-3) are constructed:

[0091]

[0092] μ i -μ j +1≤200(1-x ij ), (2)

[0093] 2≤μ i ≤201, (3)

[0094] where k = 2, ..., 201, μ i A i The order of visits in the route, i=2,...,201;

[0095] Step S435: Only the selected scenic spots can form a route, so the constraints can be constructed. Where i = 1, ..., 201;

[0096] Specifically, in step S5, the specific process of solving the model using the two-stage method is as follows:

[0097] Step S51, establishing the first stage model, the specific process of obtaining the optimal scenic spot visit set is as follows:

[0098] Step S511, the objective function is simplified to

[0099] Step S512, relax the constraints. Step S431 can be transformed into The constraint condition of step S432 can be transformed into Among them, M is an infinite number and Z is an auxiliary variable.

[0100] Step S513, using a genetic algorithm to solve the first stage model;

[0101] Step S52, establish the second stage model, and determine the specific process of the access sequence as follows:

[0102] Step S521, the objective function is simplified to

[0103] Step S522, strengthen the constraint conditions. Step S431 can be converted into Step S432 can be converted into

[0104] Step S523, the genetic algorithm solves the second stage model.

[0105] Example 2

[0106] Based on the personalized travel route planning method considering multiple topics in the aforementioned embodiment, the following is a specific example description:

[0107] Step S1, using crawler technology, obtain the evaluation data of 20 users and the data of 200 restaurants and 50 scenic spots on the Dianping website. Part of the evaluation data of the restaurant "Weixianju" is shown in Table 1;

[0108] Table 1 Partial evaluation data of the restaurant "Weixianju"

[0109]

[0110] Step S2, calculate the user's travel preferences and divide the theme tags of restaurants and scenic spots. After removing irrelevant high-frequency words, the keywords are "service", "price", "taste", "environment", "parking", "dining", "cost-effectiveness", etc.

[0111] Step S3, the interest matching scores of 20 users and 200 restaurants are combined with the fixed scores of restaurants to calculate the comprehensive scores of users for each restaurant. The entropy weight method calculates that the weights of the three indicators, local ratings, number of repeat customers, and number of check-ins in the past 30 days, are 0.4, 0.3, and 0.3, respectively. The scores obtained by this method and the scores after plain text topic matching are compared with the actual user ratings, and some results are shown in Table 2.

[0112] Table 2 Partial scoring results

[0113] Serial number Actual score This method scores Text matching score 1 5 4.6 4 2 5 4.8 4.5 3 4.5 4.3 3.5 4 4 4.1 3.5 5 5 4.7 4 6 3.5 4.0 3 7 4 4.1 3.5 8 3.5 4 2

[0114] After calculation, the comprehensive accuracy of the ratings obtained by this method reached 80%, and the accuracy of the extremely poor and excellent restaurants was 88.2% and 87.5% respectively. Compared with the traditional method, this method effectively identified user preferences by using text reviews without the need for individual user privacy information (age, gender, preferences, etc.); while the accuracy of the pure text matching method was only 73%. This shows that the fixed rating of restaurants considered by this method, on the basis of considering the matching degree of user preferences and product attributes, is conducive to further improving the accuracy. .

[0115] Step S4, the user gives the travel theme "team building" or "party", a given starting point, and combines travel time and other restrictions to establish a multi-theme restriction optimization model;

[0116] Step S5, using the two-stage method and genetic algorithm to solve the model, calculate the optimal path under 5 different starting points, and the calculation results are shown in Table 3. Taking Baita Station as an example, the optimal route under the multi-theme demand restriction is "Baita Station-Pantan Ancient Town-Shanyin Coffee (Danzhu Store)-Shenxianju Scenic Area-Xiaoshan Village Farmhouse-Baita Station", which all meet the theme requirements of "team building" and "party", and their scores are 4.8, 4.4, 4.1, and 4.7 respectively, with an average score of 4.5, and the total travel time is 65 minutes; if the theme requirements are not considered, the average score of the optimal route is slightly higher than 4.5, but only 1 meets the theme requirements. It can be concluded that the optimal solution under multi-theme restrictions can basically meet the theme requirements at the cost of reducing a small amount of scores, which meets the objectives of the present invention.

[0117] Table 3 Comparison of solutions

[0118]

[0119]

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A personalized travel route planning method considering multiple topics, characterized in that: The following steps are involved: S1, obtain user travel data and scenic spot data; S2, calculates user travel preferences and divides scenic spots into themes; S3, calculating the ratings of each user for each scenic spot; S4, building a multi-topic restricted optimization model; S5, solve the model using a two-stage approach and genetic algorithm.

2. The personalized travel route planning method considering multiple topics according to claim 1 is characterized in that: In step S1, obtaining user travel and scenic spot data specifically includes the following steps: Step S11, obtaining scenic spot data through Meituan, Dianping, and Douyin channels, and comprehensively scoring, evaluation data set PD, recommended products, and regions to form a scenic spot set A; Step S12, obtaining user travel data, region, user historical rating data R and comment text data UD through Meituan, Dianping and Douyin channels.

3. The personalized travel route planning method considering multiple topics according to claim 1 is characterized in that: In step S2, the user's travel preferences are calculated and the scenic spot themes are divided, which specifically includes the following steps: Step S21, using the jieba word segmentation package in Python to process the user comment text data UD, and using the TF-IDF algorithm to calculate the word frequency, after algorithm processing, extract s high-frequency keywords from the comment text data UD to form the user preference attribute set I= <I1,I2,……,I s >, its corresponding preference value is i= <i1,i2,……,i s >, where I s is the word to be calculated, j is the comment containing the word, j∈UD, For the I in comment j s The number of is the total number of words in comment j; Step S22, taking restaurants as an example, using jieba to perform word segmentation processing on the restaurant evaluation dataset PD, using the LDA topic model to extract the sub-topics and key words of each restaurant, analyzing the extraction results one by one, removing irrelevant words, modifying the number of topics one by one, but not modifying the number of key words; Step S23, clustering the sub-topics into four categories according to the subject words by using the K-means clustering method, and calculating the relevance to each topic. The specific process is as follows: Step S231, randomly select the initial centers of 4 classes; Step S232: In the i-th iteration, for any sample, calculate its Euclidean distance to the four centers respectively according to the word frequency, and classify the sample into the class where the center with the smallest distance is located; Step S233, using the mean method to update the center value of the class; Step S234, looping steps S232 and S233, if the center values ​​of the four cluster centers remain unchanged, the iteration ends; Step S235, according to the actual situation, comprehensively adjust and determine the theme tags, which are "environment", "service", "taste", "average price", and "party". According to the TF-IDF algorithm, the relevance of each theme word of each scenic spot in its evaluation data set is calculated, and the theme tag is K= <K1,K2,……,K n >, the corresponding correlation is k = <k1,k2,……,k n >, where k n Keyword K n 's relevance.

4. The personalized travel route planning method considering multiple topics according to claim 1 is characterized in that: In step S3, the score of each scenic spot is calculated, which specifically includes the following steps: Step S31, calculate the relationship between user u and scenic spot A i The interest matching score is: Step S32: Combine the local ratings, number of repeat visitors, and number of check-ins in the past 30 days of the scenic spot, and use the entropy weight method to calculate the fixed rating of the i-th scenic spot. The calculation formula is r i =∑ j w j ·a ij , where w j =(1+(lnn) -1 ·∑ i p ij ·lnp ij )·(∑ j w j b ij ) -1 , a ij is the value of the jth indicator of the ith scenic spot, b ij for a ij Normalized value.

5. The personalized travel route planning method considering multiple topics according to claim 1 is characterized in that: The step S4, establishing a multi-topic restriction optimization model, specifically includes the following steps: Step S41, considering the actual travel situation of the user, the following assumptions are made for the model: Step S411: The starting point of the user's travel route is A1, the set of candidate attractions is A, and the number of candidate attractions is N. A ; Step S412: If the evaluation given by each user to each scenic spot is true and reliable and reflects his or her true preferences, then user u's evaluation of scenic spot A is i Rating where r i u For user u to visit attraction A i Interest matching score; Step S42, with the goal of maximizing user satisfaction, establish the objective function as where y i ={0,1} indicates whether attraction A is selected in the route i ; Step S43: restricting the model in consideration of the user's actual travel situation.

6. The personalized travel route planning method considering multiple topics according to claim 5 is characterized in that: In step S43, the following restrictions are imposed on the model in consideration of the user's actual travel situation: Step S431: The longest playing time of the user for this trip is T max , since the traffic time is random, chance constrained programming is adopted, and the time window constraint is in v i For Attraction A i The residence time, t ij For Attraction A i With attractions A j The travel time between Step S432: for the desired topic label B given by the user, the route contains at most S max attractions that do not meet the theme requirements, then the theme constraint is in B≥1, if attraction A i If the topic tag meets the requirements, K i For Attraction A i Collection of hashtags Step S433: The user's travel route must form a loop, so a constraint is constructed. Step S434: Each scenic spot can only be visited once, so the following constraints are constructed: m i -m j +1≤(N A -1)(1-x ij ),(2) 2≤μ i ≤N A ,(3) Where k = 2, ..., N A ,μ i A i The order of visits in the route, i=2,...,N A ; Step S435: If only the selected scenic spots constitute the route, then the constraints are constructed. i where i=1,...,N A .

7. The personalized travel route planning method considering multiple topics according to claim 1 is characterized in that: The step S5, using a two-stage method to solve the model, specifically includes the following steps: Step S51, establishing a first-stage model to obtain an optimal scenic spot visit set, specifically includes the following steps: Step S511, the objective function is simplified to Step S512, relax the constraints, and convert step S431 into The constraint condition of step S432 is transformed into Among them, M is an infinite number, and Z is an auxiliary variable; Step S52, establishing a second-stage model and determining the access sequence, specifically includes the following steps: Step S521, the objective function is simplified to Step S522, strengthen the constraint conditions, and convert step S431 into Step S432 is converted into

Citation Information

Patent Citations

  • Personalized travel package recommendation method based on demand classification and subject analysis

    CN105955981A

  • Travel and trip integrated planning method considering travel demands and preferences of users

    CN113886699A

  • Route recommendation method and device and storage medium

    CN117668362A

  • Group tourist route recommendation method based on attention mechanism

    WO2022088661A1