Tourist route intelligent recommendation method and system based on scenic spot text travel fusion feature similarity
By constructing a knowledge base of scenic spot resources and utilizing multimodal feature extraction and time decay weighting methods, the recommendation of tourist routes is optimized, solving the problems of user preferences not being considered and large computational load in existing technologies, and achieving efficient and accurate recommendation of tourist resources.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SICHUAN TOURISM UNIV
- Filing Date
- 2026-02-02
- Publication Date
- 2026-05-12
AI Technical Summary
Existing tourism resource recommendation methods fail to fully consider user preferences, especially the recommendations for attractions at the boundaries of the target area are inaccurate, and the system requires a large amount of computation, increasing operating costs.
This paper proposes an intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots. By constructing a knowledge base of scenic spot resources, utilizing multimodal feature extraction and an improved time decay weighting method, and combining the historical tourism data of target users, the method calculates the similarity of scenic spots and optimizes the recommended routes, taking into account the influence of scenic spots in neighboring areas.
While ensuring the accuracy of recommendations, the server's computational load was reduced, lowering system operating costs. Furthermore, the recommended routes better match users' preferences, saving on travel expenses.
Smart Images
Figure CN122019892A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of tourism route recommendation technology, specifically relating to a method and system for intelligent tourism route recommendation based on the similarity of cultural and tourism integration characteristics of scenic spots. Background Technology
[0002] With the improvement of people's economic level and the continuous advancement of big data and network technology, the tourism industry has developed rapidly, and tourism resource recommendation software has become an important tool for enhancing users' travel experience and promoting tourism consumption. Faced with a large amount of tourism information and diverse user needs, providing personalized and accurate tourism resource recommendations to target users is particularly important.
[0003] Currently, tourism resource recommendation methods include content-based recommendation methods, user-based recommendation methods, large-scale model recommendation methods, and comprehensive recommendation methods. Among them, content-based recommendation methods study the target user's historical behavior, analyze the similarity between historical tourist attractions and other tourist attractions, and recommend attractions similar to the historical tourist attractions based on the similarity score. If the target user has a large number of historical tourist attractions, calculating the similarity between each historical tourist attraction and other tourist attractions sequentially would result in a large computational load, requiring the system to have higher requirements and better hardware, which would inevitably increase operating costs.
[0004] Existing recommendation algorithms analyze the similarity between all attractions within a destination region and historical tourist attractions to recommend attractions. However, in practice, some recommended attractions may be located on the boundary of the destination region, while a neighboring attraction may be located nearby and have a high similarity to the target user's historical tourist attractions, making it a desired destination. Existing methods do not consider this issue, and the recommendation results do not fully consider user preferences, thus lacking comprehensiveness. Summary of the Invention
[0005] To address the aforementioned technical issues, this invention provides a method and system for intelligently recommending tourist routes based on the similarity of cultural and tourism integration characteristics of scenic areas.
[0006] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: an intelligent tourism route recommendation method based on the similarity of cultural and tourism integration characteristics of scenic spots. This method obtains cultural and tourism integration characteristics based on the cultural and tourism information of scenic spots, and integrates the cultural and tourism integration characteristics of different scenic spots in different regions to construct a scenic spot resource knowledge base; it includes the following steps: S1. Based on the target user's historical travel interest information, obtain historical tourist attractions and extract the cultural and tourism integration characteristics of historical tourist attractions from the attraction resource knowledge base; S2. Obtain the cultural and tourism integration characteristics of all attractions in the target area from the attraction resource knowledge base, and select a recommendation method based on the number of historical tourist attractions: if the number of historical tourist attractions is less than or equal to a preset threshold, calculate the first similarity between historical tourist attractions and attractions in the target area, integrate all attractions in the target area and their corresponding first similarities to form a first set; if the number of historical tourist attractions is greater than the preset threshold, proceed to step S3. S3. Based on whether the target user has already visited the historical tourist attractions, the historical tourist attractions are divided into intended historical tourist attractions and visited historical tourist attractions; the web browsing records of intended historical tourist attractions are obtained, and the attention of intended historical tourist attractions is calculated based on the improved time decay weighting method. Attractions with high attention are recorded as key intended historical tourist attractions; the second similarity between key intended historical tourist attractions, visited historical tourist attractions and attractions in the target area is calculated, and all attractions in the target area and their corresponding second similarities are integrated to form a second set. S4. Select a recommended attractions with high similarity from the first set or the second set to form the third set. Determine whether there are attractions in the neighboring areas based on whether there are attractions at the boundary in the third set. Calculate the third similarity between historical tourist attractions and attractions in the neighboring areas. Compare the similarity between attractions in the neighboring areas and a recommended attractions to determine whether to update the third set. S5. Based on the dynamic constraints of the target users, plan travel routes for the a recommended attractions in the third set; S6. The target user expresses a desire not to visit the recommended attractions in the third set. This recommended attraction is recorded as a rejected attraction. The similarity of the rejected attractions in the first and second sets is then assigned to 0. The first and second sets are updated according to the new similarity ranking, and the process returns to step S4.
[0007] Preferably, the construction of the scenic spot resource knowledge base includes the following process: S01. Acquire and integrate cultural and tourism information from different regions and scenic spots. Cultural information includes remote sensing image data of regional characteristics, text data and image data of ethnic culture, and tourism information includes text data and image data of scenic spot evaluation. S02, Using two dimensions The filter algorithm extracts texture features from remote sensing image data; the Doc2vec algorithm extracts text features; and the image2vec algorithm extracts image features. After multimodal feature extraction, the triplet representation of the scenic spot's cultural tourism integration features is obtained, as follows: ; Indicates tourist attractions The cultural and tourism integration characteristics, where i represents the i-th scenic spot, and i is a positive integer;
[0008] Representing the texture features, text features, and image features of the scenic spot; S03. Convert the triplet representation into a concatenated vector and use the PCA dimensionality reduction algorithm to obtain a compact representation of the cultural tourism integration features: ; S04. Represent the cultural and tourism integration characteristics of all attractions in a region as a set, and store them in the attraction resource knowledge base according to the different regional sets.
[0009] Preferably, the number of historical tourist attractions is denoted as N, and the total number of attractions in the target area is denoted as M. In step S2, each target area attraction has N first similarity scores, and the largest first similarity score s1 is selected; the target area attraction m i and its corresponding maximum first similarity s 1mi As a data set (m i ,s 1mi Store and integrate data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the first similarity value to form a first set; m i Let i represent the i-th target area attraction, i=1,2,3,……,M.
[0010] Preferably, in step S3, each target area attraction has N second similarities, and the largest second similarity s2 is selected; the target area attraction m i and its corresponding maximum second similarity s 2mi As a data set (m i ,s 2mi Store and integrate the data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the second similarity value to form a second set.
[0011] Preferably, the web browsing records include browsing frequency, browsing duration, access depth, and interactive behavior; the method for calculating the attention score of intended historical tourist attractions in step S3 is as follows: S31. Obtain web browsing history data for all intended historical tourist attractions within a period of time, and normalize the web browsing history data. S32. Calculate the behavioral attention score of the i-th intended historical tourist attraction on the j-th day. : =ω1*PL+ω2*JS+ω3*HD+ω4*SD PL represents the frequency of visits to attractions within a day, with ω1 as its corresponding weight; JS represents the average browsing time of attractions within a day, with ω2 as its corresponding weight; HD represents the interaction rate of attractions within a day, with ω3 as its corresponding weight; SD represents the depth of visits to attractions within a day, with ω4 as its corresponding weight; ω1 + ω2 + ω3 + ω4 = 1; i = 1, 2, ..., n1, where n1 represents the number of intended historical tourist attractions; j = 1, 2, ..., L, where L represents the number of days in which the target user has a web browsing record within a period; S33. Determine the attention score of the i-th intended historical tourist attraction. : ; λ is the attenuation coefficient; t i Let j be the time difference between the j-th day and the current time, in days.
[0012] Preferably, the scenic spot at the boundary represents the area D that intersects with the neighboring area, which is centered on the current scenic spot and has a radius of r. Step S4 includes the following specific steps: S41. Select a recommended attractions with high similarity from the first set or the second set, sort the data groups of a recommended attractions according to the similarity, and form a third set; select the attraction with the lowest similarity from the third set and call it the reference attraction, and the similarity corresponding to the reference attraction is called the reference similarity. S42. Determine whether there are any scenic spots at the boundary in the third set: If no attractions exist at the boundary, the third set will not be updated; If there are attractions at the boundary, proceed to step S43; S43. Determine whether there are any neighboring attractions within area D near the attraction at the boundary: If there are no nearby attractions, the third set will not be updated; If there are attractions in neighboring areas, calculate the third similarity between the historical tourist attraction and the attractions in neighboring areas, and then determine the magnitude of the third similarity: If the third similarity is greater than the reference similarity, and there is only one reference attraction, then replace the data set of the reference attraction in the third set with an array of attractions in the neighboring region, and update the third set. If the third similarity is greater than the reference similarity, and the number of reference attractions is greater than or equal to 2, then proceed to step S44; If the third similarity is less than or equal to the reference similarity, then the third set is not updated; S44. If a reference point is also a boundary point, it is recorded as a special point; determine whether a reference point is a special point: If there are special attractions among multiple reference attractions, replace the data group of special attractions in the third set with the array of attractions in the neighboring area, and update the third set; If no special attraction exists among the multiple reference attractions, replace the array of reference attractions that is furthest from the attractions in the neighboring region with the array of attractions in the neighboring region, and update the third set.
[0013] Preferably, the target user dynamic restrictions in step S5 include travel time, departure point, mode of travel, real-time opening hours and reception capacity of attractions, and traffic conditions.
[0014] Accordingly, a smart tourism route recommendation system based on the similarity of cultural and tourism integration characteristics of scenic spots is provided. The system executes the smart tourism route recommendation method process. The system includes a user terminal, a server terminal, and a cloud terminal. The user terminal includes a human-computer interaction interface, which is used by the target user to access the system and display and query the results of analysis and calculation by the server. The cloud includes a scenic spot resource knowledge base, which stores scenic spot data from different regions and scenic spots. The scenic spot data includes the cultural and tourism integration characteristics of the scenic spots, cultural information and tourism information of the scenic spots. The server-side includes a first similarity calculation module, a second similarity calculation module, an analysis and processing module, and a third similarity calculation module; When the number of historical tourist attractions is less than or equal to a preset threshold, the first similarity calculation module is used to calculate the similarity between the historical tourist attractions and all attractions in the target area. When the number of historical tourist attractions exceeds a preset threshold, the second similarity calculation module is used to calculate the similarity between the intended historical tourist attractions, the historical tourist attractions already visited, and all attractions in the target area. When there are scenic spots at the boundary in the third set, and there are scenic spots in neighboring areas near the scenic spots at the boundary, the third similarity calculation module is used to calculate the third similarity between the historical tourist attractions and the scenic spots in neighboring areas. The analysis and processing module obtains scenic spot information from the cloud and transmits it to the first and second similarity calculation modules. Based on the calculation results of the first and second similarity calculation modules, it constructs the first, second, and third sets. It selects reference similarity and reference scenic spots from the third set. It compares the similarity between scenic spots in neighboring areas and each scenic spot in the third set to determine whether to update the third set. Based on the target user's dynamic constraints, it plans travel routes for the recommended scenic spots in the third set.
[0015] The present invention has the following beneficial effects: 1. The recommendation method of this invention constructs a scenic spot resource knowledge base based on multimodal features, using remote sensing image data of regional characteristics, textual and image data of ethnic culture, and textual and image data of scenic spot evaluation. It obtains historical tourist attractions based on the target user's historical travel interest information and selects an appropriate recommendation method based on the number of historical tourist attractions. If the number of historical tourist attractions is small, the similarity between the historical tourist attractions and all attractions in the destination area is directly calculated for recommendation. If the number of historical tourist attractions is large, the attention of the intended historical tourist attractions is calculated based on an improved time decay weighted method, selecting 3-5 attractions with high target user attention, and then calculating the similarity between these 3-5 attractions and all attractions in the destination area for recommendation. While ensuring accurate and comprehensive recommendations for the target user, this method significantly reduces server computation and lowers system operating costs.
[0016] 2. In the recommendation method of this invention, if among the selected *a* recommended attractions, there happens to be an attraction located at the boundary of the destination region, and there also happens to be an attraction in a neighboring region at the boundary of a neighboring region, and the similarity between this neighboring attraction and the target user's historical travel attractions is very high, then the attraction with the lowest similarity among the selected *a* attractions will be replaced with this neighboring attraction, updating the *a* recommended attractions in the third set, and recommendations will be made again. When the target user goes to the attraction at the boundary, since the neighboring attraction is close to the boundary attraction and is one of the attractions the user wants to visit, this recommendation method will not change much from the initially planned route, saving travel costs and conforming to the target user's wishes. Attached Figure Description
[0017] Figure 1 This is a flowchart of the intelligent travel route recommendation method of the present invention. Detailed Implementation
[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.
[0019] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.
[0020] like Figure 1 As shown, this invention discloses an intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots. It obtains cultural and tourism integration features based on the cultural and tourism information of scenic spots, and integrates these features from different regions and scenic spots to construct a scenic spot resource knowledge base; the method includes the following steps: S1. Obtain basic information and historical travel interest information of the target user. Based on the target user's historical travel interest information, integrate N historical tourist attractions and obtain the cultural and tourism integration characteristics of N historical tourist attractions from the attraction resource knowledge base. Historical travel interest information includes the target user's attraction browsing history, published travel records (including scenic spot reviews, travel notes, travel videos, etc.), or attractions already visited (including flight tickets, ticket information, location information, etc.).
[0021] S2. Obtain the cultural and tourism integration characteristics of all M attractions in the target area from the scenic spot resource knowledge base. Select a recommendation method based on the number of historical tourist attractions: If N is less than or equal to the preset threshold, calculate the first similarity between the N historical tourist attractions and the M attractions in the target area, and integrate the M attractions in the target area and their corresponding first similarities to form the first set; if N is greater than the preset threshold, proceed to step S3.
[0022] It should be understood that each target area attraction has N first similarity scores, and the largest first similarity score s1 is selected as the element of the first set. The target area attractions m... i and its corresponding maximum first similarity s 1mi As a data set (m i ,s 1mi Store and integrate data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the first similarity value to form a first set; m i Let represent the i-th target area attraction, i = 1, 2, 3, ..., M. 1mi Indicates the target area / attraction m i The highest first similarity.
[0023] S3. Based on whether the target user has already visited the attraction, divide the N historical tourist attractions into n1 intended historical tourist attractions and n2 visited historical tourist attractions, N = n1 + n2; obtain the web browsing records of the n1 intended historical tourist attractions, including browsing frequency, browsing duration, access depth, and interactive behavior; calculate the attention of the intended historical tourist attractions based on the improved time decay weighted method, sort the n1 intended historical tourist attractions in descending order of the calculated attention, and select the n3 attractions with the highest attention as key intended historical tourist attractions; calculate the second similarity between the n3 key intended historical tourist attractions, the n2 visited historical tourist attractions, and the M attractions in the target area, integrate the M attractions in the target area and their corresponding second similarities to form a second set.
[0024] It should be understood that each target area attraction has N second similarities, and the largest second similarity s2 is selected as the element of the second set. This is applied to the target area attractions m. i and its corresponding maximum second similarity s 2mi As a data set (m i ,s 2mi Store and integrate data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the second similarity value to form a second set; 2mi Indicates the target area / attraction m i The second highest similarity.
[0025] S4. Select a recommended attractions with high similarity from the first or second set to form the third set. Determine whether there are attractions in the neighboring areas based on whether there are attractions at the boundary in the third set. Calculate the third similarity between historical tourist attractions and attractions in the neighboring areas. Compare the similarity between attractions in the neighboring areas and a recommended attractions to determine whether to update the third set.
[0026] A boundary point is a scenic spot that intersects with a neighboring region D, centered on the current scenic spot and with a radius of r. A neighboring region refers to the area adjacent to the target region.
[0027] It should be noted that, in this invention, adjacent regions can be between provinces, between cities, between counties, between urban areas, between districts, between cities and counties, between provinces and counties, etc. N, M, n1, n2, n3, and a are all positive integers.
[0028] S5. Based on the dynamic constraints imposed on the target users, plan travel routes for the 'a' recommended attractions in the third set. The dynamic constraints include travel time, departure point, mode of transportation, real-time opening hours and capacity of the attractions, and traffic conditions.
[0029] S6. The target user expresses a desire not to visit a certain recommended attraction in the third set. This recommended attraction is recorded as a rejected attraction. The similarity of the rejected attractions in the first and second sets is then assigned to 0. The first and second sets are updated according to the new similarity ranking, and the process returns to step S4.
[0030] Furthermore, the construction of the scenic spot resource knowledge base includes the following process: S01. Acquire and integrate cultural and tourism information from different regions and scenic spots. Cultural information includes remote sensing image data of regional characteristics, text data and image data of ethnic culture, and tourism information includes text data and image data of scenic spot evaluation.
[0031] S02. A two-dimensional Gabor filter algorithm is used to extract texture features from remote sensing image data to reflect the common visual attributes of similar land features, represented as the following convolutional form:
[0032] ;
[0033] in, For the input remote sensing image data, This is the kernel function of a two-dimensional Gabor filter. To ensure that the generated remote sensing texture features have the same dimensionality as the text and image features, and to avoid affecting the compact representation of subsequent cultural tourism fusion features, we propose to optimize the texture features... Expand by row: ; The Doc2vec algorithm is used to extract text features. As an unsupervised model, Doc2vec can convert variable-length text data into fixed-length vector features, and can extract features from mixed text, as follows: ; in Given a sentence in a text database, its feature representation is obtained using the Doc2vec algorithm. Considering the large amount of information in the text, we will focus on scenic spots and describe them. Integrate text features: ; Image features are extracted using image2vec and expanded row by row: ; ; After completing the multimodal feature extraction, the triplet representation of the cultural and tourism integration features of scenic spots is obtained as follows: ; Indicates tourist attractions The cultural and tourism integration characteristics, where i represents the i-th scenic spot, and i is a positive integer;
[0034] Indicates tourist attractions Texture features, text features, and image features.
[0035] S03. Convert the triplet representation into a concatenated vector and use the PCA dimensionality reduction algorithm to obtain a compact representation of the cultural tourism integration features: .
[0036] S04. Represent the cultural and tourism integration characteristics of all attractions in a region as a set, and store them in the attraction resource knowledge base according to the different regional sets.
[0037] Furthermore, in step S3, the attention score of intended historical tourist attractions is calculated based on an improved time decay weighted method. This fully considers the time factor; tourist attractions whose online browsing records are more recent than the current time are considered more likely to be visited in the near future. The specific method steps are as follows: S31. Obtain web browsing history data for all intended historical tourist attractions within a given period and normalize the web browsing history data. For example, obtain the web browsing history data of the target user from January 1st to June 30th within the past six months. Specifically, obtain web browsing history data for the following five days: January 10th (the 5th day of web browsing), February 5th (the 4th day of web browsing), March 9th (the 3rd day of web browsing), March 24th (the 2nd day of web browsing), and May 8th (the 1st day of web browsing). Obtain the browsing frequency, browsing duration, access depth (number of web pages viewed, etc.), and interaction behavior (likes, comments, etc.) data for these five days, and normalize these four indicators to the range of 0-1.
[0038] S32. Calculate the behavioral attention score of the i-th intended historical tourist attraction on the j-th day. : =ω1*PL+ω2*JS+ω3*HD+ω4*SD; PL represents the frequency of visits to attractions within a day, with ω1 as its corresponding weight; JS represents the average browsing time of attractions within a day, with ω2 as its corresponding weight; HD represents the interaction rate of attractions within a day, with ω3 as its corresponding weight; SD represents the depth of visits to attractions within a day, with ω4 as its corresponding weight; ω1 + ω2 + ω3 + ω4 = 1; i = 1, 2, ..., n1, where n1 represents the number of intended historical tourist attractions; j = 1, 2, ..., L, where L represents the number of days in which the target user has a web browsing record within a period.
[0039] S33. Determine the attention score of the i-th intended historical tourist attraction. : ; λ is the decay coefficient, which controls the decay rate over time; for example, it can be 0.45. i Let j be the time difference between the j-th day and the current time, in days.
[0040] Furthermore, step S4 specifically includes the following steps: S41. Select a recommended attractions with high similarity from the first or second set, sort the data groups of a recommended attractions according to the similarity, and form a third set; select the attraction with the lowest similarity from the third set and call it the reference attraction, and the similarity corresponding to the reference attraction is called the reference similarity.
[0041] S42. Determine whether there are any scenic spots at the boundary in the third set: If no attractions exist at the boundary, the third set will not be updated; If there are attractions at the boundary, proceed to step S43.
[0042] S43. Determine whether there are any neighboring attractions within area D near the attraction at the boundary: If there are no nearby attractions, the third set will not be updated; If there are attractions in neighboring areas, calculate the third similarity between the historical tourist attraction and the attractions in neighboring areas, and then determine the magnitude of the third similarity: If the third similarity is greater than the reference similarity, and there is only one reference attraction, then replace the data set of the reference attraction in the third set with an array of attractions in the neighboring region, and update the third set. If the third similarity is greater than the reference similarity, and the number of reference attractions is greater than or equal to 2, then proceed to step S44; If the third similarity is less than or equal to the reference similarity, then the third set is not updated.
[0043] S44. If a reference point is also a boundary point, it is recorded as a special point; determine whether a reference point is a special point: If there are special attractions among multiple reference attractions, replace the data group of special attractions in the third set with the array of attractions in the neighboring area, and update the third set; If no special attraction exists among the multiple reference attractions, replace the array of reference attractions that is furthest from the attractions in the neighboring region with the array of attractions in the neighboring region, and update the third set.
[0044] The recommendation principle of the intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots disclosed in this invention is as follows: First, a scenic spot resource knowledge base is constructed based on the cultural and tourism information of the scenic spots. Then, the historical tourism attention information of the target user is obtained to determine N historical tourist attractions. If the number of historical tourist attractions is small, the similarity between the historical tourist attractions and all attractions in the destination area is directly calculated, and the a attractions with high similarity are selected for tourism route planning. If the number of historical tourist attractions is large, in order to reduce the server's computing load, the attention of the intended historical tourist attractions is calculated based on the improved time decay weighted method, and 3-5 attractions with high attention from the target user are selected. Then, the similarity between these 3-5 attractions and all attractions in the destination area is calculated, and the a attractions with high similarity are selected for tourism route planning. If one of the selected a attractions happens to be located at the boundary of the destination area, and there is also an attraction at the boundary of a neighboring area, and the similarity between this attraction and the target user's historical tourist attractions is very high, then this attraction replaces the attraction with the lowest similarity among the selected a attractions, and the recommendation is repeated. This recommendation method not only reduces server computation, but also takes into full account the nearby attractions that the target user wants to visit based on the planned route. It does not change much from the initially planned route, which not only saves travel costs, but also meets the wishes of the target user.
[0045] Specifically, this invention also discloses a smart tourism route recommendation system based on the similarity of cultural and tourism integration features of scenic spots. The system executes the smart tourism route recommendation method process; the system includes a user terminal, a server terminal, and a cloud terminal. The user terminal includes a human-computer interaction interface, which is used by the target user to access the system and display and query the results of analysis and calculation by the server. The cloud includes a scenic spot resource knowledge base, which stores scenic spot data from different regions and scenic spots. The scenic spot data includes the cultural and tourism integration characteristics of the scenic spots, cultural information and tourism information of the scenic spots. The server-side includes a first similarity calculation module, a second similarity calculation module, an analysis and processing module, and a third similarity calculation module; When the number of historical tourist attractions is less than or equal to a preset threshold, the first similarity calculation module is used to calculate the similarity between the historical tourist attractions and all attractions in the target area. When the number of historical tourist attractions exceeds a preset threshold, the second similarity calculation module is used to calculate the similarity between the intended historical tourist attractions, the historical tourist attractions already visited, and all attractions in the target area. When there are scenic spots at the boundary in the third set, and there are scenic spots in neighboring areas near the scenic spots at the boundary, the third similarity calculation module is used to calculate the third similarity between the historical tourist attractions and the scenic spots in neighboring areas. The analysis and processing module obtains scenic spot information from the cloud and transmits it to the first and second similarity calculation modules. Based on the calculation results of the first and second similarity calculation modules, it constructs the first, second, and third sets. It selects reference similarity and reference scenic spots from the third set. It compares the similarity between scenic spots in neighboring areas and each scenic spot in the third set to determine whether to update the third set. Based on the target user's dynamic constraints, it plans travel routes for the recommended scenic spots in the third set.
[0046] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any modifications, alterations, alterations, or substitutions made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention shall fall within the protection scope defined by the claims of the present invention.
Claims
1. A method for intelligently recommending tourist routes based on the similarity of cultural and tourism integration characteristics of scenic spots, characterized by: Based on the cultural and tourism information of scenic spots, we obtain the characteristics of cultural and tourism integration, and integrate the characteristics of cultural and tourism integration of different scenic spots in different regions to construct a scenic spot resource knowledge base; including the following steps: S1. Based on the target user's historical travel interest information, obtain historical tourist attractions and extract the cultural and tourism integration characteristics of historical tourist attractions from the attraction resource knowledge base; S2. Obtain the cultural and tourism integration characteristics of all attractions in the target area from the attraction resource knowledge base, and select a recommendation method based on the number of historical tourist attractions: if the number of historical tourist attractions does not exceed the preset threshold, calculate the first similarity between historical tourist attractions and attractions in the target area, integrate all attractions in the target area and their corresponding first similarities to form a first set; if the number of historical tourist attractions is greater than the preset threshold, proceed to step S3. S3. Divide historical tourist attractions into intended historical tourist attractions and visited historical tourist attractions; obtain the online browsing records of intended historical tourist attractions, calculate the attention of intended historical tourist attractions based on the improved time decay weighting method, and record the attractions with high attention as key intended historical tourist attractions; calculate the second similarity between key intended historical tourist attractions, visited historical tourist attractions and attractions in the target area, integrate all attractions in the target area and their corresponding second similarities to form a second set. S4. Select a recommended attractions with high similarity from the first set or the second set to form the third set. Determine whether there are attractions in the neighboring areas based on whether there are attractions at the boundary in the third set. Calculate the third similarity between historical tourist attractions and attractions in the neighboring areas. Compare the similarity between attractions in the neighboring areas and a recommended attractions to determine whether to update the third set. S5. Based on the target user's dynamic constraints, plan travel routes for the a recommended attractions in the third set.
2. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: The construction of the scenic spot resource knowledge base includes the following process: S01. Acquire and integrate cultural and tourism information from different regions and scenic spots. Cultural information includes remote sensing image data of regional characteristics, text data and image data of ethnic culture, and tourism information includes text data and image data of scenic spot evaluation. S02, Using two dimensions The filter algorithm extracts texture features from remote sensing image data; image2vec is used to extract image features; after multimodal feature extraction, the triplet representation of the scenic spot's cultural tourism integration features is obtained, as follows: ； Indicates tourist attractions The cultural and tourism integration characteristics, where i represents the i-th scenic spot, and i is a positive integer; Representing the texture features, text features, and image features of the scenic spot; S03. Convert the triplet representation into a concatenated vector and use the PCA dimensionality reduction algorithm to obtain a compact representation of the cultural tourism integration features: ； S04. Represent the cultural and tourism integration characteristics of all attractions in a region as a set, and store them in the attraction resource knowledge base according to the different regional sets.
3. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: The number of historical tourist attractions is denoted as N, and the total number of attractions in the target area is denoted as M. In step S2, each target area attraction has N first similarity scores, and the largest first similarity score s1 is selected; the target area attraction m i and its corresponding maximum first similarity s 1mi As a data set (m i ,s 1mi Store and integrate data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the first similarity value to form a first set; m i Let i represent the i-th target area attraction, i=1,2,3,……,M.
4. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 3, characterized in that: In step S3, each target area attraction has N second similarities, and the largest second similarity s2 is selected; the target area attractions m i and its corresponding maximum second similarity s 2mi As a data set (m i ,s 2mi Store and integrate the data sets of all attractions in the target area, and sort the data sets of all attractions in the target area according to the second similarity value to form a second set.
5. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: The web browsing history includes browsing frequency, browsing duration, access depth, and interactive behavior; the method for calculating the attention score of intended historical tourist attractions in step S3 is as follows: S31. Obtain web browsing history data for all intended historical tourist attractions within a period of time, and normalize the web browsing history data. S32. Calculate the behavioral attention score of the i-th intended historical tourist attraction on the j-th day. : =ω1*PL+ω2*JS+ω3*HD+ω4*SD; PL represents the frequency of visits to attractions within a day, with ω1 as its corresponding weight; JS represents the average browsing time of attractions within a day, with ω2 as its corresponding weight; HD represents the interaction rate of attractions within a day, with ω3 as its corresponding weight; SD represents the depth of visits to attractions within a day, with ω4 as its corresponding weight; ω1 + ω2 + ω3 + ω4 = 1; i = 1, 2, ..., n1, where n1 represents the number of intended historical tourist attractions; j = 1, 2, ..., L, where L represents the number of days in which the target user has a web browsing record within a period; S33. Determine the attention score of the i-th intended historical tourist attraction. : ； λ is the attenuation coefficient; t i Let j be the time difference between the j-th day and the current time, in days.
6. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: Boundary attractions refer to attractions that have geographical intersections with adjacent areas within a radius of r centered on the attraction. Step S4 includes the following specific steps: S41. Select a recommended attractions with high similarity from the first set or the second set, sort the data groups of a recommended attractions according to the similarity, and form a third set; select the attraction with the lowest similarity from the third set and call it the reference attraction, and the similarity corresponding to the reference attraction is called the reference similarity. S42. Determine whether there are any scenic spots at the boundary in the third set: If no attractions exist at the boundary, the third set will not be updated; If there are attractions at the boundary, proceed to step S43; S43. Determine whether there are any neighboring attractions within area D near the attraction at the boundary: If there are no nearby attractions, the third set will not be updated; If there are attractions in neighboring areas, calculate the third similarity between the historical tourist attraction and the attractions in neighboring areas, and then determine the magnitude of the third similarity: If the third similarity is greater than the reference similarity, and there is only one reference attraction, then replace the data set of the reference attraction in the third set with an array of attractions in the neighboring region, and update the third set. If the third similarity is greater than the reference similarity, and the number of reference attractions is greater than or equal to 2, then proceed to step S44; If the third similarity is less than or equal to the reference similarity, then the third set is not updated; S44. If a reference point is also a boundary point, it is recorded as a special point; determine whether a reference point is a special point: If there are special attractions among multiple reference attractions, replace the data group of special attractions in the third set with the array of attractions in the neighboring area, and update the third set; If no special attraction exists among multiple reference points, replace the reference point that is furthest from the nearby attractions with that nearby attraction and update the third set.
7. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: The target user dynamic restrictions in step S5 include travel time, departure point, mode of travel, real-time opening hours and reception capacity of attractions, and traffic conditions.
8. The intelligent tourism route recommendation method based on the similarity of cultural and tourism integration features of scenic spots according to claim 1, characterized in that: It also includes step S6, where the target user expresses a desire not to visit the recommended attractions in the third set, and this recommended attraction is recorded as a rejected attraction. Then, the similarity of the rejected attractions in the first and second sets is assigned to 0, the first and second sets are updated according to the new similarity ranking, and the process proceeds to step S4 again.
9. A smart tourism route recommendation system based on the similarity of cultural and tourism integration characteristics of scenic spots, characterized by: The system is used to execute the intelligent travel route recommendation method according to any one of claims 1 to 8; the system includes a user terminal, a server terminal, and a cloud terminal; The user terminal includes a human-computer interaction interface, which is used by the target user to access the system and display and query the results of analysis and calculation by the server. The cloud includes a scenic spot resource knowledge base, which stores scenic spot data from different regions and scenic spots. The scenic spot data includes the cultural and tourism integration characteristics of the scenic spots, cultural information and tourism information of the scenic spots. The server-side includes a first similarity calculation module, a second similarity calculation module, an analysis and processing module, and a third similarity calculation module; When the number of historical tourist attractions is less than or equal to a preset threshold, the first similarity calculation module is used to calculate the similarity between the historical tourist attractions and all attractions in the target area. When the number of historical tourist attractions exceeds a preset threshold, the second similarity calculation module is used to calculate the similarity between the intended historical tourist attractions, the historical tourist attractions already visited, and all attractions in the target area. When there are scenic spots at the boundary in the third set, and there are scenic spots in neighboring areas near the scenic spots at the boundary, the third similarity calculation module is used to calculate the third similarity between the historical tourist attractions and the scenic spots in neighboring areas. The analysis and processing module obtains scenic spot information from the cloud and transmits it to the first and second similarity calculation modules. Based on the calculation results of the first and second similarity calculation modules, it constructs the first, second, and third sets. It selects reference similarity and reference scenic spots from the third set. It compares the similarity between scenic spots in neighboring areas and each scenic spot in the third set to determine whether to update the third set. Based on the target user's dynamic constraints, it plans travel routes for the recommended scenic spots in the third set.