Tourism customization method and device and storage medium

Abstract

The invention discloses a travel customization method and device and a storage medium, and relates to the technical field of travel information services, and the method comprises the steps: obtaining a basic travel constraint condition inputted by a user through a terminal; guiding the user to submit multi-modal data based on the basic travel constraint condition, and performing deep analysis on the multi-modal data by using a multi-modal large model to generate a user deep preference vector; based on the basic travel constraint condition and the user depth preference vector, screening out a candidate resource set from a travel resource database; accessing a real-time data source through a data interface to obtain dynamic environment information, and constructing a personalized space-time reachability dynamic graph for a candidate resource set in combination with the user depth preference vector and the dynamic environment information; and running a multi-objective optimization algorithm on the personalized space-time reachability dynamic graph. The method effectively solves the problems that a traditional method cannot deeply understand user preferences and lacks dynamic response ability.

Description

Technical Field

[0001] This invention relates to the field of tourism information service technology, and in particular to a method, apparatus and storage medium for customized tourism services. Background Technology

[0002] With the deep integration of information technology and the tourism industry, customized travel services have become a key development direction in the online travel sector. Traditional customized travel solutions are mostly based on rule engines and collaborative filtering technologies. For example, they build user profiles using historical user behavior data or explicit ratings, and then match and recommend destinations using a static resource library. These solutions typically employ a modular combination of flights and hotels, or offer limited themed route options, which to some extent meet users' demand for standardized travel products. Existing technologies can now perform preliminary resource screening by analyzing users' basic travel constraints and improve the accuracy of the booking process by accessing some real-time data through API interfaces.

[0003] However, the above methods still have significant limitations in terms of personalized experience and dynamic adaptability. Most existing systems rely on historical data and explicit feedback to build user preference models, making it difficult to capture users' deep-seated and multimodal potential needs. Traditional text analysis cannot effectively process this type of unstructured data, resulting in coarse-grained user profiles lacking emotional dimensions. It usually treats itinerary planning as a static resource sequence recommendation problem, failing to dynamically couple real-time environmental data with users' personalized preferences. This leads to the generated travel plans lacking robustness in real-world scenarios and being unable to cope with itinerary adjustment needs under unexpected circumstances, putting users at risk of experience gaps and planning failures during actual travel. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a travel customization method that solves the problem of insufficient robustness in itinerary planning caused by the failure of existing technologies to dynamically couple real-time environmental data with user preferences.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for customizing travel, which includes obtaining basic travel constraints input by a user through a terminal;

[0008] Based on basic travel constraints, users are guided to submit multimodal data, and a large multimodal model is used to perform in-depth analysis on the multimodal data to generate user deep preference vectors.

[0009] Based on basic travel constraints and user deep preference vectors, a set of candidate resources is selected from the tourism resource database;

[0010] accessing a real-time data source through a data interface to obtain dynamic environment information, and constructing a personalized spatiotemporal reachability dynamic graph for the candidate resource set in combination with the user deep preference vector and the dynamic environment information;

[0011] running a multi-objective optimization algorithm on the personalized spatiotemporal reachability dynamic graph to optimize trip matching degree, time efficiency and cost control targets, and generating a tourism plan with anti-interference capability;

[0012] outputting the anti-interference tourism plan through a human-computer interaction interface, and executing reservation of the tourism plan.

[0013] As a preferred scheme of the tourism customization method, the method comprises the following steps:

[0014] An application program on the terminal device shows a structured data input interface to the user;

[0015] The user inputs information in the mandatory fields in the structured input interface presented by the terminal according to the guidance of the field labels, which together constitute the complete basic travel constraint condition.

[0016] As a preferred scheme of the tourism customization method, the method comprises the following steps:

[0017] Through self-supervised learning on a large-scale image-text pair data set, alignment representation between visual and text modalities and cross-modality semantic association are learned, and a pre-trained multi-modal large model is obtained;

[0018] The destination information in the basic travel constraint condition is input into the pre-trained multi-modal large model, the multi-modal large model analyzes the destination information, and generates a topic word distribution vector;

[0019] The user encapsulates the visual examples and text prompts provided by the multi-modal data collection interface into multi-modal data triples, extracts frame sequences through frame sampling, and processes the description text through word segmentation and cleaning to form standardized multi-modal data triples;

[0020] The standardized multi-modal data triples are input into the pre-trained multi-modal large model, the multi-modal large model extracts a picture deep visual feature vector and a video overall representation vector through an internal visual encoder, extracts a text semantic feature vector through a text encoder, fuses the visual and text features through an internal cross-modality attention computing mechanism, generates aligned visual context vectors and text context vectors, and splices and fuses to form a multi-modal joint feature vector.

[0021] The multimodal joint feature vector is mapped to the fully connected layer for dimension mapping. The fully connected layer then uses the Sigmoid activation function to convert the multimodal joint feature vector into a user deep preference vector.

[0022] As a preferred embodiment of the tourism customization method described in this invention, the following steps are included: Based on basic travel constraints and user deep preference vectors, a candidate resource set is selected from a tourism resource database:

[0023] Each tourism resource unit in the tourism resource database is extracted using a multimodal large model to generate a tourism resource feature vector;

[0024] The destination's geographical range and budget limit in the basic travel constraints are used to traverse the tourism resource database, excluding tourism resource units whose geographical location exceeds the destination's geographical range and whose price exceeds the budget limit, thus forming a preliminary filtered subset of tourism resource units.

[0025] Based on the user's deep preference vector and the tourism resource feature vector of each tourism resource unit in the initially filtered tourism resource unit subset, a cosine similarity is calculated to generate a matching score for each tourism resource unit.

[0026] Based on the preliminary filtering of tourism resource units, each tourism resource unit is sorted from high to low according to its matching score, and the top N tourism resource units are selected to form a candidate resource set.

[0027] As a preferred embodiment of the tourism customization method of the present invention, the method involves: obtaining dynamic environmental information by accessing a real-time data source through a data interface, and constructing a personalized spatiotemporal reachability dynamic graph for the candidate resource set by combining the user's deep preference vector and the dynamic environmental information, including the following steps:

[0028] Dynamic environmental information is obtained by accessing real-time data sources through data interfaces and integrated into a standardized dynamic environmental information vector set;

[0029] Based on the candidate resource set and dynamic environmental information vector set, a bidirectional edge is established between two tourism resource unit nodes. The initial attributes of the bidirectional edge are set with basic geographical distance and transportation mode options, forming a spatiotemporal graph static skeleton.

[0030] The dynamic environment information vector set is input into the spatiotemporal graph static skeleton. Based on the real-time traffic conditions in the dynamic environment information vector set, the path planning method based on real-time traffic big data is used to obtain the movement time cost of each edge.

[0031] According to the queuing data in the dynamic environment information vector set, the queuing prediction rule of real-time bearing capacity analysis is used to obtain the waiting time cost of each node, and a basic space-time reachability graph is generated;

[0032] Based on the user deep preference vector and the basic space-time reachability graph, the dimensions in the user deep preference vector are adjusted, the comprehensive cost of the edges is calculated, and a personalized space-time reachability dynamic graph is generated.

[0033] As a preferred scheme of the travel customization method, a multi-objective optimization algorithm is run on the personalized space-time reachability dynamic graph, the travel matching degree, time efficiency and cost control target are optimized, and a travel scheme with anti-interference ability is generated, including the following steps:

[0034] A multi-objective optimization function of the travel matching degree, time efficiency and cost control target is defined using a Pareto optimization method on the personalized space-time reachability dynamic graph;

[0035] A genetic algorithm is used to solve the multi-objective optimization function on the personalized space-time reachability dynamic graph, and an initial population is evolved by iteratively performing selection, crossover and mutation operations on the personalized space-time reachability dynamic graph, and the optimal path is obtained by selecting the individual with the highest fitness;

[0036] When the dynamic environment information changes, an anti-interference re-planning mechanism is triggered, a local path re-optimization is performed on the personalized space-time reachability dynamic graph based on the current user location and the remaining candidate resource set, and an adjusted path is obtained.

[0037] The optimal path and the adjusted path are combined to generate an anti-interference travel scheme.

[0038] As a preferred scheme of the travel customization method, the anti-interference travel scheme is output through a human-computer interaction interface, and the reservation of the travel scheme is performed, including the following steps:

[0039] The anti-interference travel scheme is presented to the user in the form of a visual calendar through the human-computer interaction interface, and the reservation process is triggered after the user confirms the anti-interference travel scheme through the human-computer interaction interface;

[0040] The reservation process calls a standardized data interface to send the resource unit information in the anti-interference travel scheme to a third-party service access end;

[0041] The third-party service access end returns reservation confirmation information and displays the operation execution result on the human-computer interaction interface.

[0042] In a second aspect, the present application provides a travel customization device, comprising an acquisition module, which acquires the basic travel constraint conditions input by the user through a terminal;

[0043] The analysis module guides the user to submit multi-modal data based on the basic travel constraint condition, and uses a multi-modal large model to deeply analyze the multi-modal data, and generates a user deep preference vector;

[0044] The screening module screens a candidate resource set from a tourism resource database based on the basic travel constraint condition and the user deep preference vector;

[0045] The construction module accesses a real-time data source through a data interface to obtain dynamic environment information, and constructs a personalized spatio-temporal reachability dynamic graph for the candidate resource set in combination with the user deep preference vector and the dynamic environment information.

[0046] The tourism scheme generation module runs a multi-objective optimization algorithm on the personalized spatio-temporal reachability dynamic graph, optimizes travel matching degree, time efficiency and cost control targets, and generates a travel scheme with anti-interference capability.

[0047] The execution module outputs the anti-interference travel scheme through a human-computer interaction interface, and executes the reservation of the travel scheme.

[0048] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the travel customization method according to the first aspect of the present application is implemented.

[0049] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, any step of the travel customization method according to the first aspect of the present application is implemented.

[0050] The present application has the following beneficial effects: a multi-modal large model is used to deeply analyze the picture, video and text data submitted by the user to generate a precise user deep preference vector, a personalized spatio-temporal reachability dynamic graph that integrates user preferences is constructed in combination with real-time dynamic environment information, and then a multi-objective optimization algorithm is used to generate a travel scheme with anti-interference capability on the graph, thereby realizing intelligent travel customization with high personalization, strong dynamic adaptability and the ability to respond to real-time changes, and effectively solving the problems that the traditional method cannot deeply understand user preferences and lacks dynamic response capability. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0052] Fig. 1Flowchart of a travel customization method.

[0053] Fig. 2 Schematic diagram of a file encryption system.

[0054] Fig. 3 Schematic diagram of a structured data input interface.

[0055] Fig. 4 Flowchart of a user depth preference vector. DETAILED DESCRIPTION

[0056] In order to make the above objectives, characteristics and advantages of the present application more obvious and understandable, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0057] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0058] Secondly, the "one embodiment" or "embodiment" referred to herein means that a specific feature, structure or characteristic can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0059] REFERENCE Figs. 1-4 For one embodiment of the present application, the embodiment provides a travel customization method, comprising the following steps:

[0060] S1, obtaining the basic travel constraint conditions input by the user through the terminal.

[0061] S1.1, the application program on the terminal device shows the user a structured data input interface.

[0062] Further, the application program on the terminal device shows the user a structured data input interface, which presents a plurality of input fields with explicit labels in a form layout, the field types include a date selector for collecting travel dates, a geographic search box for inputting destination cities or regions, a digital input box for filling in the total number of travel days, and a currency input box for setting the maximum budget amount. The structured interface design effectively avoids confusion and omission during user input through clear visual guidance and logical grouping, ensuring complete collection of the required information of the basic travel constraint conditions.

[0063] S1.2, the user inputs information in the structured input interface presented by the terminal according to the guidance of the field label, and the information of the mandatory fields constitutes a complete basic travel constraint condition.

[0064] Further, the user inputs information in the structured input interface presented by the terminal according to the guidance of the field label, selects or inputs the travel time period in the date selector, enters or selects the destination name from the drop-down list in the geographic search box, inputs or adjusts the travel duration through the keyboard or the stepper in the numerical input box, and inputs or sets the upper limit of the budget amount through the slider in the currency input box. After the information of all mandatory fields is input, a structured data object containing time, place, duration and budget attributes is encapsulated, that is, a complete basic travel constraint condition, which effectively prevents incomplete information or format errors through mandatory item inspection and real-time format verification, ensuring the data quality and availability of the basic travel constraint condition.

[0065] S2, based on the basic travel constraint condition, guiding the user to submit multi-modal data, and using a multi-modal large model to deeply analyze the multi-modal data to generate a user deep preference vector.

[0066] S2.1, by self-supervised learning on a large-scale image-text pair data set, learning the alignment representation and cross-modal semantic association between visual and text modalities, obtaining a pre-trained multi-modal large model.

[0067] Further, by training the multi-modal large model on a large-scale image-text pair data set using self-supervised learning, the multi-modal large model learns the alignment representation and cross-modal semantic association between visual and text modalities using mask language modeling and image completion tasks, and finally obtains a pre-trained multi-modal large model. This method fully utilizes unannotated image-text data to mine deep semantic associations through self-supervised pre-training, effectively overcoming the problem of excessive dependence on annotated data in traditional supervised learning.

[0068] S2.2, input the destination information in the basic travel constraint condition into the pre-trained multi-modal large model, and the multi-modal large model analyzes the destination information to generate a topic word distribution vector.

[0069] Further, the destination information text in the basic travel constraint condition is input into the pre-trained multi-modal large model, the multi-modal large model extracts semantic features of the destination text through an internal text encoder, and generates a topic word distribution vector associated with visual concepts using a cross-modal mapping layer. Each dimension weight in the topic word distribution vector represents the association strength of the destination with a specific topic. Through the semantic understanding ability of the multi-modal large model, abstract destination information is converted into quantifiable topic preference guidance.

[0070] S2.3, the user encapsulates the visual examples and text prompts provided by the multi-modal data collection interface into multi-modal data triples, extracts frame sequences through frame sampling, and describes the text after word segmentation and cleaning to form standardized multi-modal data triples.

[0071] Further, the user selects and submits personal collection of travel pictures, short video clips and text descriptions according to the visual examples and text prompts provided by the multi-modal data collection interface. These multi-modal data are encapsulated into multi-modal data triples containing image, video and text elements. The video data is extracted into key frame sequences through equal interval frame sampling, and the text description is cleaned through word segmentation and stop word filtering. Finally, the standardized multi-modal data triples are formed. The standardized processing procedure ensures the quality of multi-modal data through unified data format and cleaning specification.

[0072] S2.4, the standardized multi-modal data triples are input into the pre-trained multi-modal large model. The multi-modal large model extracts image deep visual feature vectors and video overall representation vectors through internal visual encoders, and extracts text semantic feature vectors through text encoders. The multi-modal large model fuses visual and text features through the built-in cross-modal attention mechanism to generate aligned visual context vectors and text context vectors, and then splices and fuses them to form multi-modal joint feature vectors.

[0073] Further, the standardized multi-modal data triples are input into the pre-trained multi-modal large model. The multi-modal large model extracts image deep visual feature vectors and video overall representation vectors through internal visual encoders, and extracts text semantic feature vectors through text encoders. The multi-modal large model calculates the correlation weight between the text semantic feature vectors and the visual feature vectors through the built-in cross-modal attention mechanism, generates aligned visual context vectors and text context vectors, and splices and fuses all feature vectors to form multi-modal joint feature vectors. Through the cross-modal attention mechanism, the deep semantic fusion of visual and text features is realized, and the complex semantic association between multi-modal data is effectively captured.

[0074] S2.5, the multi-modal joint feature vector is mapped to the full connection layer for dimension mapping. The full connection layer converts the multi-modal joint feature vector into a user deep preference vector through the Sigmoid activation function.

[0075] Further, the multi-modal joint feature vector is input into a fully connected layer for linear transformation, and the fully connected layer maps the multi-modal joint feature vector to a preset dimension preference space through a Sigmoid activation function. The mapping output is the user deep preference vector, and each element in the user deep preference vector has a value range between zero and one. The numerical size directly quantifies the user's tendency strength for the corresponding preference dimension. This step converts the fused multi-modal features into an interpretable user preference quantitative representation through dimension mapping of the fully connected layer and the Sigmoid activation function.

[0076] S3, based on the basic travel constraint condition and the user deep preference vector, filtering out a candidate resource set from the tourism resource database.

[0077] S3.1, extracting each tourism resource unit in the tourism resource database through the multi-modal large model to generate a tourism resource feature vector.

[0078] Further, each tourism resource unit in the tourism resource database is processed by the multi-modal large model. The multi-modal large model uses an internal visual encoder to analyze the landscape pictures of the tourism resource unit to extract deep visual features, uses a text encoder to analyze the description text of the tourism resource unit to extract semantic features, and fuses these features to generate a tourism resource feature vector. This method converts unstructured tourism resource information into a standardized numerical vector through the powerful representation capability of the multi-modal large model, effectively solving the problem that traditional keyword matching cannot capture deep features of resources.

[0079] S3.2, the destination geographical range and the budget upper limit in the basic travel constraint condition, traversing the tourism resource database, excluding tourism resource units whose geographical position exceeds the destination geographical range, and excluding tourism resource units whose price exceeds the budget upper limit, to form a preliminary filtered subset of tourism resource units.

[0080] Further, the destination geographical range and the budget upper limit in the basic travel constraint condition are used as filtering conditions to traverse all tourism resource units in the tourism resource database, exclude tourism resource units whose latitude and longitude coordinates exceed the polygon boundary of the destination geographical range, and exclude tourism resource units whose price value exceeds the budget upper limit value. The remaining tourism resource units that meet the conditions form a preliminary filtered subset of tourism resource units. Through geographical space filtering and economic constraint filtering, the candidate range is quickly narrowed, effectively reducing the subsequent calculation complexity and ensuring the feasibility of the scheme.

[0081] S3.3, based on the user deep preference vector and the tourism resource feature vector of each tourism resource unit in the preliminary filtered subset of tourism resource units, performing cosine similarity calculation to generate a matching score for each tourism resource unit.

[0082] The similarity expression is:

[0083] ;

[0084] in, For the first The deep preference vector of the i-th user and the th The matching score of the similarity between the feature vectors of each tourism resource. For the first A user's deep preference vector For the first The th user's deep preference vector Element values ​​in each dimension Index for users, An index for tourism resource units. For the first Feature vector of a tourism resource unit The total number of dimensions for user preferences and tourism resource characteristics. Indexed by user preferences and tourism resource characteristics. For the first The feature vector of the tourism resource unit is... The element values ​​of each dimension.

[0085] Furthermore, cosine similarity is calculated based on the user's deep preference vector and the tourism resource feature vector of each tourism resource unit in the initially filtered tourism resource unit subset. The calculation process uses the mathematical formula of vector dot product divided by the product of the magnitudes to generate a matching score between zero and one for each tourism resource unit. The higher the matching score, the closer the direction of the tourism resource feature vector is to the user's deep preference vector, i.e., the higher the degree of matching. The cosine similarity measure achieves accurate quantitative matching between user preferences and resource characteristics, overcoming the subjectivity and one-sidedness of traditional rule matching.

[0086] S3.4 Based on the preliminary filtering of tourism resource units, each tourism resource unit is sorted from high to low according to the matching score, and the top N tourism resource units are selected to form a candidate resource set.

[0087] Furthermore, based on the preliminary filtered tourism resource unit subset, each tourism resource unit is sorted from high to low according to its matching score. The top N tourism resource units with the highest matching scores in the sorted list are selected to form a candidate resource set. The candidate resource set contains N tourism resource unit identifiers and their attribute information that best match the user's deep preference vector. The screening strategy ensures that the candidate resources not only meet the basic constraints but are also highly relevant to the user's preferences by sorting and selecting the Top-N. This provides a high-quality and personalized resource pool for subsequent itinerary planning.

[0088] S4. Obtain dynamic environment information by accessing real-time data sources through data interfaces, and construct a personalized spatiotemporal reachability dynamic graph for the candidate resource set by combining user deep preference vectors and dynamic environment information.

[0089] S4.1. Obtain dynamic environment information by accessing real-time data sources through data interfaces and integrate it into a standardized dynamic environment information vector set.

[0090] Furthermore, dynamic environmental information is obtained by accessing real-time data sources through data interfaces. These real-time data sources include traffic data interfaces that provide real-time road condition predictions and travel times, location data interfaces that provide the number of people queuing at scenic spots and the estimated waiting time, and weather data interfaces that provide precipitation probability and temperature impact coefficients. This information is integrated into a standardized dynamic environmental information vector set, which includes timestamps and geographic coordinate labels to ensure data timeliness and spatial relevance. Through multi-source real-time data fusion and standardized processing, the problem that static data cannot reflect dynamic environmental changes is solved.

[0091] S4.2 Based on the candidate resource set and dynamic environmental information vector set, a bidirectional edge is established between two tourism resource unit nodes. The initial attributes of the bidirectional edge are set with basic geographical distance and transportation mode options, forming a spatiotemporal graph static skeleton.

[0092] Furthermore, based on the candidate resource set and dynamic environmental information vector set, each tourism resource unit in the candidate resource set is mapped as a graph node. The node attributes include the geographical coordinates and resource type of the tourism resource unit. A bidirectional edge is established between any two tourism resource unit nodes. The initial attributes of the bidirectional edge are set with basic geographical distance and optional transportation options, forming a spatiotemporal graph static skeleton containing nodes and edges. Discrete tourism resource units are connected into a network through graph structure modeling.

[0093] S4.3 Input the dynamic environment information vector set into the spatiotemporal graph static skeleton, and obtain the movement time cost of each edge based on the real-time traffic conditions in the dynamic environment information vector set and the path planning method based on real-time traffic big data.

[0094] Furthermore, the dynamic environment information vector set is input into the spatiotemporal graph static skeleton. Based on the real-time traffic data in the dynamic environment information vector set, the path planning method based on real-time traffic big data calculates the movement time cost of each edge. The movement time cost includes traffic time cost and transfer time cost. The traffic time cost predicts the travel time of different modes of transportation based on real-time traffic conditions, and the transfer time cost considers the walking and waiting time for public transportation transfers. By injecting real-time traffic data, the static skeleton is transformed into a dynamic network, which accurately reflects the changes in time consumption during the actual movement process.

[0095] S4.4 Based on the queuing data in the dynamic environment information vector set, the waiting time cost of each node is obtained using the queuing prediction rules of real-time capacity analysis, and a basic spatiotemporal reachability map is generated.

[0096] Furthermore, based on the queuing data in the dynamic environment information vector set, the queuing prediction rule based on real-time capacity analysis is used to calculate the waiting time cost of each node. The queuing prediction rule predicts the waiting time in the future time period based on the real-time number of people queuing and the venue capacity. The waiting time cost is added as a node attribute to the static skeleton of the spatiotemporal graph to generate a basic spatiotemporal reachability graph that includes dynamic movement costs and waiting costs.

[0097] S4.5. Based on the fusion of user deep preference vector and basic spatiotemporal reachability graph, the weights of each dimension in the user deep preference vector are adjusted and the comprehensive cost of the edge is calculated to generate a personalized spatiotemporal reachability dynamic graph.

[0098] The comprehensive cost expression is:

[0099] ;

[0100] in, Starting node To the target node The final overall cost, Starting node To the target node Basic mobility costs, Based on the weighting coefficient, For preference dimension The adjustment coefficient, For users in the first The intensity of dimensional preference, For the first Correlation between user preference dimensions Starting node To the target node Edge objects.

[0101] Furthermore, based on the fusion of user deep preference vectors and basic spatiotemporal accessibility graphs, the weights of each dimension in the user deep preference vectors are used to adjust the comprehensive cost calculation of edges. The preference dimension weights affect the cost weights of different modes of transportation through adjustment coefficients. For example, a high comfort preference weight will increase the comprehensive cost of edges with more transfers, and a high economic preference weight will increase the comprehensive cost of expensive modes of transportation. Finally, a personalized spatiotemporal accessibility dynamic graph reflecting individual preferences is generated.

[0102] S5. Run a multi-objective optimization algorithm on a personalized spatiotemporal accessibility dynamic graph to optimize the objectives of itinerary matching, time efficiency, and cost control, and generate a travel plan with strong anti-interference capabilities.

[0103] S5.1. Using the Pareto optimization method on a personalized spatiotemporal accessibility dynamic graph, define a multi-objective optimization function with objectives of trip matching degree, time efficiency, and cost control.

[0104] Furthermore, a multi-objective optimization function is defined using the Pareto optimization method on the personalized spatiotemporal accessibility dynamic graph. The multi-objective optimization function includes the trip matching degree objective function, which calculates the average cosine similarity between all nodes in the path and the user's deep preference vector; the time efficiency objective function, which calculates the weighted sum of the total travel time and total waiting time of the path; and the cost control objective function, which calculates the deviation of the total cost of the path from the budget limit. The objective functions are combined into a comprehensive fitness function through a weighted summation method. By considering multiple dimensions such as matching degree, time and cost simultaneously through the multi-objective optimization framework, the bias problem of the solution caused by single-objective optimization is avoided.

[0105] S5.2. A genetic algorithm is used to solve the multi-objective optimization function on the personalized spatiotemporal reachability dynamic graph. The initial population is evolved by iteratively performing selection, crossover and mutation operations on the personalized spatiotemporal reachability dynamic graph. The optimal path is obtained by selecting the individual that converges to the one with the highest fitness.

[0106] The overall fitness value is expressed as:

[0107] ;

[0108] in, The overall fitness value, For user deep preference vectors, For the first Tourism resource units in each location, For the first The price of a tourism resource unit at a given location As the weight for trip matching degree, Weighted by time efficiency, As a weight for cost control, This represents the total number of location nodes for tourism resource units in the previous candidate paths. This serves as the location node index for tourism resource units.

[0109] Furthermore, a genetic algorithm is used to solve a multi-objective optimization function on a personalized spatiotemporal reachability dynamic graph. The genetic algorithm encodes the path as a chromosome sequence, and the chromosome gene sequence represents the access order of tourism resource units. The initial population is generated by random walks on the personalized spatiotemporal reachability dynamic graph. The genetic algorithm retains individuals with high fitness through roulette wheel selection, generates new individuals by exchanging gene segments of parent chromosomes through sequential crossover, and changes the gene sequence order through random exchange mutation. During the iterative evolution process, the average comprehensive fitness of each generation of the population continuously improves. Finally, the individual with the highest fitness is selected to decode the optimal path. This optimization process efficiently searches for the global optimum in a complex solution space by simulating the natural evolution mechanism, overcoming the limitation of traditional planning algorithms that are prone to getting trapped in local optima.

[0110] S5.3 When dynamic environmental information changes are detected, an anti-interference replanning mechanism is triggered. Based on the current user location and the remaining candidate resource set, local path re-optimization is performed on the personalized spatiotemporal reachability dynamic graph to obtain the adjusted path.

[0111] Furthermore, when dynamic environmental information changes are detected, an anti-interference replanning mechanism is triggered. Based on the current user location coordinates and the remaining candidate resource set, the current location is used as the new path starting point on the personalized spatiotemporal reachability dynamic graph, and the unvisited candidate resource set is used as the new node pool. The genetic algorithm is re-run to optimize the local path and obtain an adjusted path that adapts to the latest environmental conditions. By responding to environmental changes in real time through dynamic planning, the problem of fixed itinerary plans failing when faced with emergencies is effectively solved, ensuring the real-time nature and feasibility of the travel plan.

[0112] S5.4. Merge the optimal path and the adjusted path to generate an interference-resistant tourism plan.

[0113] Furthermore, the optimal path obtained by the genetic algorithm is merged with the adjusted path obtained by the anti-interference replanning mechanism. The merging process retains the actual visit records of the executed path nodes, and the unvisited node sequence of the adjusted path is connected to the current visited node to generate a complete anti-interference tourism plan that includes historical visited nodes and future adjusted nodes. The merging operation achieves a smooth connection between historical itineraries and future plans through path splicing technology, generating an anti-interference tourism plan.

[0114] S6. Output an anti-interference travel plan through the human-computer interaction interface and execute the booking of the travel plan.

[0115] S6.1 The anti-interference travel plan is presented to users in the form of a visual schedule through a human-computer interaction interface. After confirming the anti-interference travel plan through the human-computer interaction interface, the user triggers the booking process.

[0116] Furthermore, the anti-interference travel plan is presented to users in the form of a visual schedule through a human-computer interaction interface. The visual schedule uses a timeline layout to display the daily itinerary. Each time node is marked with the name of the tourism resource unit, image preview, and activity duration. Real-time transportation route map and cost details are displayed below the timeline. Users can trigger the booking process by clicking the confirmation button through the human-computer interaction interface. This visual presentation method effectively solves the problems of limited information capacity and lack of intuitiveness of traditional text itineraries through intuitive timeline design and multimedia element integration, improving users' understanding efficiency and decision-making confidence in complex itinerary plans.

[0117] S6.2 The booking process calls the standardized data interface to send the resource unit information in the anti-interference tourism plan to the third-party service access terminal.

[0118] Furthermore, the booking process calls a standardized data interface to send resource unit information from the anti-interference tourism plan to the third-party service access point. The standardized data interface uses the RESTful API protocol to transmit JSON format data. The data packet contains tourism resource unit identifiers, service timestamps, user identity identifiers, and booking quantity parameters. After receiving the data, the third-party service access point parses it and forwards it to the corresponding tourism resource supplier service platform. This standardized interface design effectively solves the protocol compatibility problem when connecting multiple platforms by unifying the data format and transmission protocol, and realizes seamless data exchange with third-party services.

[0119] S6.3 The third-party service access terminal returns the reservation confirmation information and displays the operation execution result on the human-computer interaction interface.

[0120] Furthermore, the third-party service access point returns booking confirmation information and displays the operation execution result on the human-computer interaction interface. The returned booking confirmation information includes a booking success status code, resource unit confirmation number, final price details, and supplier contact information. The human-computer interaction interface displays the booking success information with a green success prompt box, while marking abnormal situations that require user attention with a red warning box. This result feedback mechanism effectively avoids the problem of opaque status in the traditional booking process through status code parsing and visual prompts, ensuring that users can keep track of the booking execution status in a timely manner.

[0121] This embodiment also provides a travel customization device, including: an acquisition module for acquiring basic travel constraints input by the user through a terminal;

[0122] The analysis module, based on basic travel constraints, guides users to submit multimodal data and uses a large multimodal model to perform in-depth analysis of the multimodal data, generating user deep preference vectors.

[0123] The filtering module selects a set of candidate resources from the tourism resource database based on basic travel constraints and user deep preference vectors.

[0124] The module connects to a real-time data source through a data interface to obtain dynamic environment information. Combining the user's deep preference vector and the dynamic environment information, it constructs a personalized spatiotemporal reachability dynamic graph for the candidate resource set.

[0125] The travel itinerary generation module runs a multi-objective optimization algorithm on a personalized spatiotemporal accessibility dynamic map to optimize itinerary matching, time efficiency, and cost control objectives, generating travel itineraries with strong anti-interference capabilities.

[0126] The execution module outputs an anti-interference travel plan through a human-computer interaction interface and executes the booking of the travel plan.

[0127] This embodiment also provides a computer device applicable to the tourism customization method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the tourism customization method proposed in the above embodiment.

[0128] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0129] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the travel customization method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0130] In summary, this invention generates accurate user deep preference vectors by deeply analyzing user-submitted image, video, and text data using a multimodal large model. It then constructs a personalized spatiotemporal accessibility dynamic graph that integrates user preferences by combining real-time dynamic environmental information. Subsequently, a multi-objective optimization algorithm is used to generate travel plans with anti-interference capabilities on this graph. This achieves highly personalized, dynamically adaptable, and real-time-responsive intelligent tourism customization, effectively solving the problems of traditional methods that cannot deeply understand user preferences and lack dynamic response capabilities.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method of customizing travel, characterized by: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

2. The travel customization method of claim 1, wherein: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

3. The travel customization method of claim 2, wherein: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

4. The travel customization method of claim 3, wherein: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the The destination geographical range and the upper limit of the budget in the basic travel constraint condition, traversing the travel resource database, excluding the travel resource units whose geographical positions exceed the destination geographical range, and excluding the travel resource units whose prices exceed the upper limit of the budget, to form a preliminary filtered subset of travel resource units; Based on the user deep preference vector and the travel resource feature vector of each travel resource unit in the preliminary filtered subset of travel resource units, cosine similarity calculation is performed to generate a matching score for each travel resource unit; Based on the preliminary filtered subset of travel resource units, each travel resource unit is sorted in descending order of matching score, and the top N travel resource units are selected to form a candidate resource set.

5. The travel customization method of claim 4, wherein: Through a data interface, real-time data sources are accessed to obtain dynamic environmental information, and the user deep preference vector and the dynamic environmental information are combined to construct a personalized spatio-temporal reachability dynamic graph for the candidate resource set, including the following steps: Through a data interface, real-time data sources are accessed to obtain dynamic environmental information, and the user deep preference vector and the dynamic environmental information are combined to construct a personalized spatio-temporal reachability dynamic graph for the candidate resource set, including the following steps: Based on the candidate resource set and the dynamic environmental information vector set, a bidirectional edge is established between two travel resource unit nodes, and the initial attributes of the bidirectional edge are set as the basic geographical distance and the transportation mode option, forming a spatio-temporal graph static skeleton; The dynamic environmental information vector set is input into the spatio-temporal graph static skeleton, and based on the real-time traffic data path planning method, the moving time cost of each edge is obtained according to the real-time traffic conditions in the dynamic environmental information vector set; According to the queuing data in the dynamic environmental information vector set, the waiting time cost of each node is obtained using the queuing prediction rules of real-time carrying capacity analysis, and a basic spatio-temporal reachability graph is generated; Based on the user deep preference vector and the basic spatio-temporal reachability graph, the dimensions of the user deep preference vector are adjusted, and the comprehensive cost of the edge is calculated to generate a personalized spatio-temporal reachability dynamic graph.

6. The travel customization method of claim 5, wherein: On the personalized spatio-temporal reachability dynamic graph, a multi-objective optimization algorithm is run to optimize the trip matching degree, time efficiency and cost control target, and a anti-interference tourism scheme is generated, including the following steps: On the personalized spatio-temporal reachability dynamic graph, a Pareto optimization method is used to define a multi-objective optimization function of the trip matching degree, time efficiency and cost control target; A genetic algorithm is used to solve the multi-objective optimization function on the personalized spatio-temporal reachability dynamic graph, and the initial population is evolved through iteration of selection, crossover and mutation operations on the personalized spatio-temporal reachability dynamic graph to obtain the optimal path by selecting the individual with the highest fitness. When the dynamic environmental information changes, an anti-interference re-planning mechanism is triggered, and based on the current user location and the remaining candidate resource set, a local path re-optimization is performed on the personalized spatio-temporal reachability dynamic graph to obtain an adjusted path. The optimal path and the adjusted path are combined to generate an anti-interference tourism scheme.

7. The travel customization method of claim 6, wherein: Through a human-computer interaction interface, the anti-interference tourism scheme is output, and the reservation of the tourism scheme is executed, including the following steps: The anti-interference tourism scheme is presented to the user in the form of a visual schedule through a human-computer interaction interface, and the user confirms the anti-interference tourism scheme through the human-computer interaction interface to trigger the reservation process; The reservation process calls a standardized data interface to send resource unit information in the anti-interference travel plan to a third-party service access end; The third-party service access end returns reservation confirmation information and displays operation execution results on a human-computer interaction interface.

8. A travel customization apparatus based on the travel customization method according to any one of claims 1 to 7, characterized by: The method comprises the following steps: a user inputs basic travel constraint conditions through a terminal; An analysis module guides the user to submit multi-modal data based on the basic travel constraint conditions, and uses a multi-modal large model to deeply analyze the multi-modal data to generate a user deep preference vector; A screening module screens a candidate resource set from a tourism resource database based on the basic travel constraint conditions and the user deep preference vector; A construction module accesses real-time data sources through a data interface to obtain dynamic environment information, and constructs a personalized spatio-temporal accessibility dynamic graph for the candidate resource set in combination with the user deep preference vector and the dynamic environment information; A travel plan generation module runs a multi-objective optimization algorithm on the personalized spatio-temporal accessibility dynamic graph to optimize travel matching degree, time efficiency, and cost control targets, and generates an anti-interference travel plan; An execution module outputs the anti-interference travel plan through a human-computer interaction interface and executes the reservation of the travel plan. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the travel customization method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the travel customization method of any one of claims 1-7.

Images