Intelligent multi-dimensional recommendation and dynamic combination system for cultural and tourism resources

Abstract

The present application belongs to the technical field, and discloses a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources; by introducing a physical preference curve and a dynamic fatigue adjustment mechanism, accurate perception of the physical state of tourists at different stages of travel and travel optimization are realized. The present application includes six key modules: user demand acquisition, physical parameter calculation, resource matching and screening, initial resource combination, fatigue balance optimization, and path planning and recommendation output. The present application can dynamically adjust the combination of cultural and travel resources according to the interests and preferences of tourists and their physical state, arrange high-intensity core scenic spots when the tourists have sufficient physical strength, and arrange relaxing experience activities during the period of physical decline. The present application also takes into account the fatigue accumulation effect and individual differences, realizes scientific and balanced fatigue throughout the journey, and improves the travel experience and satisfaction of tourists.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field, more particularly, the present application relates to a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources. BACKGROUND

[0002] With the rapid development of tourism and the increasing demand for personalization, intelligent recommendation systems for cultural and travel resources have become an important tool for enhancing the travel experience of tourists. Currently, the recommendation technology for cultural and travel resources is mainly based on collaborative filtering, content matching, and knowledge graph, etc. By analyzing the historical behavior data and interest preferences of tourists, personalized attraction recommendations and itinerary planning services are provided for users.

[0003] However, the existing recommendation systems for cultural and travel resources have a significant fatigue balancing problem when dealing with multi-day itinerary planning. These systems usually simply distribute attractions evenly across each day of the itinerary based on interest matching degree, completely ignoring the dynamic changes in the physical state of tourists during the trip. In actual travel, tourists are often energetic and in a good state on the first day of the trip, and can withstand high-intensity sightseeing activities. However, as the trip progresses, especially in the middle stage, the physical strength of tourists gradually decreases, and the accumulation of fatigue increases, and the perception of sightseeing also decreases. However, the existing system may still arrange high-intensity and long-time sightseeing in the middle and later stages when the tourist's physical strength is at its lowest, and even arrange the core attractions with the highest interest matching degree but also the highest physical consumption in this period, resulting in the inability of tourists to experience these important attractions in the best state, which seriously affects the travel satisfaction. In addition, traditional recommendation systems also lack consideration of the cumulative effect of fatigue over consecutive days, and fail to incorporate the fatigue residual factors of the previous day into the itinerary planning of the next day, and do not design differentiated fatigue threshold and recovery models for tourists of different ages and health conditions, causing a serious disconnection between itinerary arrangement and actual experience of tourists, especially for older tourists or family tourists with children, the middle and later stages of the itinerary often become a "fatigue check-in" rather than a pleasant experience.

[0004] In view of this, the present application proposes a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources, comprising:

[0006] a user demand acquisition module for acquiring itinerary planning demand information of tourists, the itinerary planning demand information including total number of days, maximum sightseeing time per day, interest preference vector, and physical preference curve;

[0007] The physical parameter calculation module is configured to obtain a daily basic fatigue threshold and a dynamic fatigue adjustment coefficient in the total number of days according to the physical preference curve, wherein the basic fatigue threshold is used to represent the maximum fatigue that the tourist can withstand in a day, and the dynamic fatigue adjustment coefficient is used to represent the amplification or reduction of the fatigue perception of the tourist due to the change in physical strength in a day.

[0008] The resource matching and screening module is configured to obtain a candidate set of cultural and tourism resources matched with the interest preference of the tourist according to the interest preference vector and a feature vector of the cultural and tourism resources in the database of cultural and tourism resources.

[0009] The initial resource combination module is configured to obtain an initial combination of cultural and tourism resources in a day in the total number of days according to the resource fatigue consumption value, the resource tour duration and the resource geographical location of each cultural and tourism resource in the candidate set of cultural and tourism resources, the maximum tour duration in a day and the basic fatigue threshold in a day.

[0010] The fatigue balance optimization module is configured to obtain an actual fatigue distribution of a daily tour according to the resource fatigue consumption value and the dynamic fatigue adjustment coefficient of the cultural and tourism resources in the initial combination of cultural and tourism resources in a day, and dynamically adjust the initial combination of cultural and tourism resources based on the actual fatigue distribution and the basic fatigue threshold to obtain an optimized combination of cultural and tourism resources with balanced fatigue.

[0011] The path planning and recommendation output module is configured to obtain a path planning result of a daily tour according to the resource geographical location and the resource tour duration of the cultural and tourism resources in the optimized combination of cultural and tourism resources in a day, and output a final tour recommendation scheme based on the path planning result and the interest preference vector.

[0012] Preferably, the step of obtaining the basic fatigue threshold and the dynamic fatigue adjustment coefficient in a day in the total number of days according to the physical preference curve comprises:

[0013] The physical preference curve is segmented to obtain a corresponding physical preference value in a day in the total number of days, wherein the physical preference value is used to represent the degree of physical strength of the tourist in the corresponding day.

[0014] The basic fatigue threshold in a day is obtained according to the product of the physical preference value and a preset fatigue reference value.

[0015] The trend of the physical preference curve in the total number of days is counted to obtain a daily physical strength decline slope and a daily physical strength fluctuation amplitude of the physical preference curve.

[0016] The basic adjustment factor in a day is obtained according to the ratio of the daily physical strength decline slope to a preset slope threshold, and the fluctuation adjustment factor in a day is obtained according to the ratio of the daily physical strength fluctuation amplitude to a preset fluctuation threshold.

[0017] The base adjustment factor and the weighted sum of the fluctuation adjustment factors are normalized to obtain a dynamic fatigue adjustment coefficient, wherein the weights of the weighted sum are determined by the age and health status of the tourists.

[0018] Preferably, the step of obtaining the candidate set of cultural and travel resources matching the interest preference of the tourists according to the interest preference vector and the feature vector of the cultural and travel resources in the database of cultural and travel resources comprises:

[0019] The resource feature vectors of each cultural and travel resource in the database of cultural and travel resources are subjected to cluster analysis to obtain a plurality of cultural and travel resource category clusters;

[0020] The Euclidean distance between the interest preference vector and the cluster center of each cultural and travel resource category cluster is calculated to obtain the top K cultural and travel resource category clusters with the highest matching degree with the interest preference vector, wherein K is a preset category cluster quantity threshold;

[0021] In the top K cultural and travel resource category clusters, the cosine similarity between the resource feature vector of each cultural and travel resource and the interest preference vector is calculated respectively, and the cultural and travel resources with a cosine similarity greater than a preset similarity threshold are recorded as initial candidate cultural and travel resources;

[0022] According to the resource geographic location of the initial candidate cultural and travel resources and the travel starting location in the travel planning requirement information, a geographic accessibility score of each initial candidate cultural and travel resource is obtained; and the initial candidate cultural and travel resources are screened based on the geographic accessibility score to obtain a candidate set of cultural and travel resources.

[0023] Preferably, the step of obtaining an initial cultural and travel resource combination for each day within the total number of days of travel according to the resource fatigue consumption value, the resource tour duration and the resource geographic location of each cultural and travel resource in the candidate set of cultural and travel resources, in combination with the maximum daily tour duration and the daily base fatigue threshold comprises:

[0024] According to the resource tour duration and the resource geographic location of each cultural and travel resource in the candidate set of cultural and travel resources, the path time consumption between each cultural and travel resource is calculated;

[0025] According to the path time consumption and the resource tour duration, a time sequence dependency graph between the cultural and travel resources is constructed, wherein the nodes of the time sequence dependency graph are the cultural and travel resources, and the edges are the path time consumption between the cultural and travel resources;

[0026] In the time sequence dependency graph, based on the maximum daily tour duration, a graph partitioning algorithm with time window constraint is adopted to divide the candidate set of cultural and travel resources into an initial subset for each day within the total number of days of travel;

[0027] For each travel resource in the daily initial subset, calculate the cumulative sum of the resource fatigue consumption value, denoted as the daily initial fatigue; compare the daily initial fatigue with the daily basic fatigue threshold value, if the daily initial fatigue is greater than the basic fatigue threshold value, remove the travel resource with the highest resource fatigue consumption value in the daily initial subset, and recalculate the daily initial fatigue, until the daily initial fatigue is less than or equal to the basic fatigue threshold value, and obtain the initial travel resource combination.

[0028] Preferably, according to the resource fatigue consumption value of each daily travel resource in the initial travel resource combination and the dynamic fatigue adjustment coefficient, the step of obtaining the actual fatigue distribution of the daily itinerary includes:

[0029] The resource fatigue consumption value of each daily travel resource in the initial travel resource combination is time-weighted, wherein the weight of time-weighting is determined by the visiting order of the travel resource in the daily itinerary, and the later the visiting order is, the higher the weight is;

[0030] The product of the time-weighted resource fatigue consumption value and the dynamic fatigue adjustment coefficient is denoted as the weighted fatigue of the daily itinerary;

[0031] According to the weighted fatigue of the daily itinerary, a fatigue distribution curve within the total number of days of the itinerary is constructed;

[0032] The fatigue distribution curve is smoothed to obtain the actual fatigue distribution.

[0033] Preferably, based on the actual fatigue distribution and the basic fatigue threshold value, the step of dynamically adjusting the initial travel resource combination to obtain the fatigue-balanced optimized travel resource combination includes:

[0034] Calculate the difference between the weighted fatigue of the daily itinerary in the actual fatigue distribution and the basic fatigue threshold value, denoted as the daily fatigue over-standard value;

[0035] If the daily fatigue over-standard value is greater than zero, identify the travel resource with the highest resource fatigue consumption value in the initial travel resource combination corresponding to the day, denoted as the to-be-adjusted travel resource;

[0036] In the candidate travel resource set, search for a replacement travel resource whose interest matching degree with the to-be-adjusted travel resource is greater than a preset matching degree threshold value and whose resource fatigue consumption value is lower than that of the to-be-adjusted travel resource;

[0037] According to the resource visiting time length and the resource geographic location of the replacement travel resource, determine whether it meets the time window constraint of the time sequence dependency graph corresponding to the day; if it meets, replace the to-be-adjusted travel resource with the replacement travel resource, and update the initial travel resource combination;

[0038] The above steps are repeated until the weighted fatigue degree of all days in the actual fatigue degree distribution is less than or equal to the basic fatigue degree threshold, and an optimized travel resource combination is obtained.

[0039] Preferably, according to the resource geographical position and resource tour duration of each daily travel resource in the optimized travel resource combination, the step of obtaining the path planning result of each daily itinerary comprises:

[0040] The resource geographical position of each daily travel resource in the optimized travel resource combination is spatially clustered to obtain a geographical partition of each daily itinerary;

[0041] In each geographical partition, an optimal tour order in the partition is constructed according to the resource tour duration and the path time consumption between travel resources, wherein the optimal tour order is determined by using a traveling salesman problem solving algorithm based on dynamic programming;

[0042] According to the spatial adjacency relationship of the geographical partitions, a partition tour order of each daily itinerary is obtained; and based on the partition tour order and the optimal tour order in the partition, a path planning result is generated.

[0043] Preferably, according to the resource geographical position of the initial candidate travel resource and the itinerary starting position in the itinerary planning requirement information, the step of obtaining the geographical accessibility score of each initial candidate travel resource comprises:

[0044] The straight-line distance between the resource geographical position of the initial candidate travel resource and the itinerary starting position is calculated, denoted as an initial distance;

[0045] According to the resource geographical position of the initial candidate travel resource, the traffic convenience degree of the region where it is located is obtained, wherein the traffic convenience degree is determined by the public transportation station density and road density around the resource geographical position;

[0046] The initial distance is negatively correlated mapped to obtain a distance score; and the distance score and the weighted sum of the traffic convenience degree are normalized to obtain the geographical accessibility score, wherein the weight of the weighted sum is determined by the transportation tool preference of the tourist.

[0047] Preferably, the fatigue degree distribution curve is smoothed, comprising:

[0048] A Gaussian kernel function is constructed, wherein the standard deviation of the Gaussian kernel function is determined by the total number of days of the itinerary and the daily physical fluctuation amplitude of the physical preference curve;

[0049] The fatigue degree distribution curve is discretized into a daily fatigue degree sequence;

[0050] The daily fatigue degree sequence and the Gaussian kernel function are convolved to obtain a smoothed fatigue degree sequence;

[0051] The smoothed fatigue degree sequence is interpolated to restore a continuous curve, denoted as an actual fatigue degree distribution.

[0052] Preferably, the step of outputting the final travel recommendation scheme based on the path planning result and the interest preference vector comprises:

[0053] According to the path planning result, a time sequence tour plan of the daily travel is generated, wherein the time sequence tour plan comprises a tour order, a predicted tour duration and a path time consumption of each daily travel resource;

[0054] According to the interest preference vector, an interest matching degree of each travel resource in the optimized travel resource combination is obtained;

[0055] The interest matching degree is associated with the time sequence tour plan to generate a recommendation reason for the daily travel, wherein the recommendation reason comprises an interest matching degree ranking of the travel resource and a resource characteristic description;

[0056] The time sequence tour plan and the recommendation reason are integrated into a visual travel recommendation scheme and output to the user terminal.

[0057] The technical effects and advantages of the multi-dimensional intelligent recommendation and dynamic combination system for travel resources of the present application are as follows:

[0058] The present application can accurately perceive the physical state changes of tourists at different stages of travel, and integrate the natural law that tourists gradually become tired from the initial full energy of the travel into the recommendation, so that the travel arrangement is more in line with the human physiological rhythm. The present application dynamically adjusts the daily travel resource combination according to the physical preference curve of the tourists, arranges higher intensity core scenic spot tours when the tourists are full of energy, and appropriately arranges relaxing and comfortable experience activities during the period of physical decline, thereby improving the perception experience of high matching degree scenic spots and the overall travel satisfaction of tourists. At the same time, the present application can also intelligently customize personalized fatigue degree thresholds and recovery models according to individual differences such as the age and health status of tourists, so that the travel planning is more inclusive and adaptive, especially suitable for diversified travel scenarios such as elderly tourists and family travel. Through the time sequence weighting and smoothing processing of fatigue degree, the present application can also effectively simulate the fatigue accumulation effect and recovery mechanism, realize the scientific balance of the whole travel fatigue degree, and enable the tourists to maintain a good state throughout the travel process and fully enjoy the travel pleasure. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The present application is a multi-dimensional intelligent recommendation and dynamic combination system for travel resources. DETAILED DESCRIPTION

[0060] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort are within the scope of the present application.

[0061] The examples of the present application provide a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources. The execution subject of the system includes but is not limited to the following: a tourism recommendation platform, an intelligent tour guide terminal, a tourism route planning system, a tourism big data analysis center, and the like, which can be regarded as general computing nodes of the present application, and the intelligent recommendation system includes but is not limited to at least one of the following: a cloud-based tourism recommendation engine, a distributed tourism resource scheduling system, and an intelligent tourism route generator.

[0062] The present application provides a multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources. By acquiring real-time travel planning demand information, physical preference curves, and interest preference vectors of tourists, an accurate daily fatigue threshold and a cultural and travel resource matching model are constructed. The optimized cultural and travel resource combination with balanced fatigue is generated by combining resource fatigue consumption values and geographic location information. Through dynamic monitoring and parameter adjustment, the present application realizes accurate matching and efficient combination of cultural and travel resources. The present application has high adaptability and can optimize travel parameters in real time according to individual differences and physical characteristics of tourists, significantly improving the tourism experience and reducing the fatigue of tourists.

[0063] Please refer to Figure 1 In the embodiments of the present application, the multi-dimensional intelligent recommendation and dynamic combination system for cultural and travel resources includes:

[0064] A user demand acquisition module is used to acquire travel planning demand information of tourists. The travel planning demand information includes total travel days, maximum daily tour duration, interest preference vectors, and physical preference curves, and the like, which are key indicators and are collected in real time through a user interaction interface. The total travel days and the maximum daily tour duration directly reflect the time constraints of tourists. The interest preference vectors represent the degree of preference of tourists for different types of cultural and travel resources. The physical preference curves describe the physical change trend of tourists during the travel period. These data provide a basis for the development of travel planning strategies, ensuring the pertinence and effectiveness of the recommendations.

[0065] The physical parameter calculation module is configured to obtain a daily basic fatigue threshold and a dynamic fatigue adjustment coefficient in the total number of days according to the physical preference curve. The basic fatigue threshold quantifies the maximum fatigue that the tourist can bear each day, directly determining the upper limit constraint of the combination of the cultural and tourism resources. The dynamic fatigue adjustment coefficient reflects the amplification or reduction of the fatigue perception of the tourist due to the change in physical condition each day, providing a key parameter for subsequent fatigue optimization. The two indexes are obtained through the segmented processing and trend analysis of the physical preference curve, ensuring that the itinerary planning matches the actual physical condition of the tourist.

[0066] The resource matching and screening module is configured to obtain a candidate set of cultural and tourism resources matching the interest preference of the tourist according to the interest preference vector and the feature vector of the cultural and tourism resources in the database. The module first performs cluster analysis on the database of cultural and tourism resources to identify the resource category cluster most relevant to the interest of the tourist, and then screens the cultural and tourism resources that best meet the needs of the tourist and are easy to reach through accurate similarity calculation and geographical accessibility evaluation, forming a candidate resource set to lay the foundation for subsequent combination optimization.

[0067] The initial resource combination module is configured to obtain an initial combination of cultural and tourism resources each day in the total number of days according to the resource fatigue consumption value, resource tour duration, and resource geographical location of each cultural and tourism resource in the candidate set of cultural and tourism resources, in combination with the maximum tour duration each day and the daily basic fatigue threshold. The module divides the candidate resource set into an initial subset each day by constructing a time sequence dependency graph among the cultural and tourism resources and using a graph partitioning algorithm with time window constraints, and optimizes and adjusts the initial subset according to the fatigue constraint, ensuring that the daily itinerary meets the tour duration requirement and does not exceed the physical load of the tourist.

[0068] The fatigue balance optimization module is configured to obtain the actual fatigue distribution of the daily itinerary according to the resource fatigue consumption value and the dynamic fatigue adjustment coefficient of each cultural and tourism resource in the initial combination of cultural and tourism resources, and dynamically adjust the initial combination of cultural and tourism resources based on the actual fatigue distribution and the basic fatigue threshold to obtain an optimized combination of cultural and tourism resources with balanced fatigue. The module accurately evaluates the fatigue distribution in the itinerary through time sequence weighting and smoothing processing of the resource fatigue, and optimizes the resources for replacement for the days exceeding the threshold, achieving balanced control of the fatigue throughout the journey and improving the tourism experience.

[0069] The path planning and recommendation output module is configured to obtain the path planning result of the daily itinerary according to the resource geographical location and resource tour duration of each cultural and tourism resource in the optimized combination of cultural and tourism resources, and output the final itinerary recommendation scheme based on the path planning result and the interest preference vector. The module determines the optimal tour order and path through spatial clustering and the traveling salesman problem solving algorithm, and generates personalized recommendation reasons in combination with the interest matching degree, finally forming an intuitive and detailed visual itinerary scheme to meet the tourism needs of the tourist.

[0070] The above various modules are connected through wired and / or wireless mode, realizing data transmission between modules.

[0071] In the embodiment of the present application, the detailed implementation steps of obtaining the basic fatigue threshold and the dynamic fatigue adjustment coefficient of each day within the total number of days of the trip according to the physical preference curve include:

[0072] The physical preference curve is segmented and processed to obtain the corresponding physical preference value of each day within the total number of days of the trip, wherein the physical preference value is used to represent the degree of physical fitness of the tourist on the corresponding day. The segmentation processing eliminates the complexity of the continuous curve, discretizes the physical state into specific numerical values of each day, and facilitates subsequent calculation. The processing process adopts an equal time interval sampling method, determines the sampling points according to the total number of days of the trip, and each sampling point corresponds to the physical state of one day. For abnormal fluctuations in the curve, a median filtering technique is used for smoothing processing to ensure the representativeness and reliability of the sampling data. The physical preference value is usually in the range of [0, 10], and the higher the value, the more vigorous the physical fitness.

[0073] The basic fatigue threshold of each day is obtained according to the product of the physical preference value and the preset fatigue reference value. The basic fatigue threshold is the upper limit of the fatigue that the tourist can bear each day, and directly affects the number and intensity of resource combinations. The calculation formula is:

[0074] ; wherein, is the basic fatigue threshold of the day, is the preset fatigue reference value, is the physical preference value of the day.

[0075] The preset fatigue reference value is a standardized parameter determined according to a large amount of user data and expert experience, and is usually set to 50-100 units, which can be dynamically adjusted according to the age, health status and other factors of the tourist. This linear mapping method is simple and intuitive, ensuring that the fatigue threshold is proportional to the physical state, and the more vigorous the physical fitness, the higher the fatigue that can be tolerated.

[0076] The trend of the physical preference curve within the total number of days of the trip is analyzed to obtain the daily physical decline slope and the daily physical fluctuation amplitude of the physical preference curve. Trend analysis is a key step to obtain the dynamic adjustment factor, which reflects the dynamic characteristics of physical changes. The daily physical decline slope is calculated by the difference between the physical preference values of adjacent two days, representing the rate of physical consumption; the daily physical fluctuation amplitude is determined by the difference between local extreme points, representing the instability degree of physical state. These two indicators together describe the dynamic characteristics of physical changes, providing a basis for subsequent adjustment coefficient calculation.

[0077] The daily basic adjustment factor is obtained according to the ratio of the daily physical decline slope to the preset slope threshold value, and the daily fluctuation adjustment factor is obtained according to the ratio of the daily physical fluctuation amplitude to the preset fluctuation threshold value. The two adjustment factors quantify the influence degree of physical change on fatigue perception. The calculation formulas are respectively:

[0078] ; wherein, is the basic adjustment factor of the day, is the physical decline slope of the day, is the preset slope threshold value.

[0079] ; wherein, is the fluctuation adjustment factor of the day, is the physical fluctuation amplitude of the day, is the preset fluctuation threshold value.

[0080] The preset slope threshold value and the preset fluctuation threshold value are reference values determined based on human physiology research and user experience data, and respectively represent the standard degree of normal physical decline and fluctuation. This ratio calculation method standardizes the original index, which is convenient for subsequent weighted fusion.

[0081] The weighted sum of the basic adjustment factor and the fluctuation adjustment factor is normalized to obtain the dynamic fatigue degree adjustment coefficient, wherein the weight of the weighted sum is determined by the age and health status of the visitor. The dynamic fatigue degree adjustment coefficient is a key parameter affecting the actual fatigue degree perception, and reflects the amplification or reduction effect of physical change. The calculation formula is:

[0082] ; wherein, is the dynamic fatigue degree adjustment coefficient of the day, is a normalization function, and are the weights of the basic adjustment factor and the fluctuation adjustment factor respectively, and .

[0083] The weight setting is dynamically adjusted according to the visitor characteristics. The older the visitor is, the higher the weight is, indicating that the influence of physical decline rate on fatigue perception is more significant; the worse the health status is, the higher the weight is, indicating that the influence of physical fluctuation on fatigue perception is more obvious. The normalization processing controls the adjustment coefficient within a reasonable range, usually [0.7, 1.3], less than 1 indicates that the fatigue perception is weakened, and greater than 1 indicates that the fatigue perception is enhanced. This personalized dynamic adjustment mechanism ensures the adaptability of the system to different visitor groups.

[0084] In the embodiments of the present application, the detailed implementation steps for obtaining a candidate set of cultural and tourism resources matching the interest preferences of tourists from the interest preference vector and the cultural and tourism resource feature vector in the cultural and tourism resource database include:

[0085] The resource feature vectors of each cultural and tourism resource in the cultural and tourism resource database are subjected to cluster analysis to obtain a plurality of cultural and tourism resource category clusters. Cluster analysis is a key step to reduce search complexity. Mass resources are classified into meaningful category clusters for quick filtering. The analysis process uses an improved K-means++ algorithm to calculate the similarity between resources through the Euclidean distance of feature vectors and iteratively optimize the division of category clusters. The clustering results usually form 20-50 different resource category clusters, each representing a category of cultural and tourism resources with similar features, such as historical sites, natural scenery, and cultural experiences. The number of category clusters is dynamically adjusted according to the size and diversity of the resource library to ensure high similarity within the category and obvious differences between categories.

[0086] The Euclidean distance between the interest preference vector and the cluster center of each cultural and tourism resource category cluster is calculated to obtain the top K cultural and tourism resource category clusters with the highest matching degree with the interest preference vector, wherein K is a preset category cluster number threshold. This step realizes coarse-grained resource filtering to quickly locate the resource category that is most likely to meet the interests of tourists. The Euclidean distance calculation uses standardization processing to eliminate the dimensional differences of different dimensional features to ensure the comparability of the calculation results. The smaller the distance value, the higher the matching degree. The preset category cluster number threshold K is usually set to 5-10 and is dynamically adjusted according to the total number of days of travel and the diversity requirements to ensure the reasonableness and sufficiency of the filtering range.

[0087] In the top K cultural and tourism resource category clusters, the cosine similarity between the resource feature vector of each cultural and tourism resource and the interest preference vector is calculated, and the cultural and tourism resources with a cosine similarity greater than a preset similarity threshold are recorded as initial candidate cultural and tourism resources. This step realizes fine-grained resource matching to accurately assess the degree of fit between each resource and the interests of tourists. The cosine similarity calculation formula is:

[0088] ; wherein, is the resource feature vector, is the interest preference vector, is the angle between the two vectors.

[0089] The cosine similarity value ranges from -1 to 1, and the closer the value is to 1, the more similar it is. The preset similarity threshold is usually set to 0.7-0.85 and is dynamically adjusted according to the clarity of the interests of tourists. The more clear the interests are, the higher the threshold is, and the more stringent the filtering is. This two-stage filtering strategy (clustering first and then similarity) greatly improves the matching efficiency while ensuring the matching accuracy.

[0090] According to the resource geographical position of the initial candidate travel resource and the trip starting position in the trip planning demand information, the geographical accessibility score of each initial candidate travel resource is obtained; the initial candidate travel resource is screened based on the geographical accessibility score, and a candidate travel resource set is obtained. The geographical accessibility evaluation is a key step to ensure the practicability of the trip, avoiding recommending resources that are difficult to reach although they match the interest. The evaluation process considers the distance factor and traffic convenience, and reflects the accessibility difficulty of the resource through comprehensive scoring. The geographical accessibility score is usually in the range of [0, 1], and the higher the value, the easier to reach. The screening process sets a dynamic threshold, which is determined according to the trip characteristics and the traffic preference of tourists, and usually selects the top 60-70% of resources to ensure that the final candidate set not only meets the interest matching requirement, but also has good accessibility.

[0091] In the embodiment of the application, according to the resource fatigue consumption value, the resource tour time and the resource geographical position of each travel resource in the candidate travel resource set, combined with the daily maximum tour time and the daily basic fatigue threshold, the detailed implementation steps of obtaining the initial travel resource combination of each day within the total number of days of the trip include:

[0092] According to the resource tour time and the resource geographical position of each travel resource in the candidate travel resource set, the path time consumption between each travel resource is calculated. The path time consumption is the basic data for planning a reasonable trip, reflecting the time cost required for transferring between resources. The calculation process is based on geographic information systems and traffic databases, considering actual road networks, traffic modes and time period factors, and obtaining accurate estimated time consumption through path planning algorithms. The calculation result forms an N1xN1 path time consumption matrix (N1 is the number of candidate resources), and each element in the matrix represents the estimated time consumption from resource k to resource j, which provides a key parameter for subsequent time sequence dependence analysis.

[0093] According to the path time consumption and the resource tour time, a time sequence dependence graph between travel resources is constructed, wherein the nodes of the time sequence dependence graph are travel resources, and the edges are the path time consumption between travel resources. The time sequence dependence graph is a special directed weighted graph, which directly shows the spatial and temporal relationship between resources. The construction process is based on a graph theory model, which takes each resource as a node of the graph, and the path time consumption between resources as the weight of the edge, and also adds the resource tour time attribute to each node. The construction of the graph considers the time window constraints such as opening time, to ensure that the model meets the actual tour conditions. The time sequence dependence graph not only reflects the physical distance between resources, but also embodies the time cost, providing an intuitive basis for decision-making for subsequent resource allocation.

[0094] In the time-dependent graph, based on the daily maximum tour duration, the graph partition algorithm with time window constraint is used to divide the candidate travel resource set into the initial daily subset within the total number of days. The graph partition is the core step of the trip day planning, and the goal is to reasonably allocate resources to each day, both to ensure that the daily tour time does not exceed the limit, and to ensure the balance of the overall resource allocation. The partition algorithm is based on the improved spectral clustering method, and the optimization partition is realized by minimizing the cutting cost of the graph. The algorithm design considers the following constraint conditions: the total time (tour duration + path time) of each day does not exceed the maximum tour duration of each day; the path time between adjacent resources is minimized as much as possible; the diversity of daily tour experience is maximized. The partition result is the preliminary daily resource allocation scheme, which provides the basis for fatigue assessment and adjustment.

[0095] For each travel resource in the initial daily subset, the cumulative sum of the resource fatigue consumption value is calculated, denoted as the initial daily fatigue. The initial daily fatigue is compared with the basic fatigue threshold of each day. If the initial daily fatigue is greater than the basic fatigue threshold, the travel resource with the highest resource fatigue consumption value is removed from the initial daily subset, and the initial daily fatigue is recalculated until the initial daily fatigue is less than or equal to the basic fatigue threshold, and the initial travel resource combination is obtained. Fatigue assessment and adjustment is a key step to ensure the comfort of the trip, and through iterative optimization, the daily fatigue is ensured not to exceed the bearing capacity of the tourists. The evaluation process first accumulates the fatigue consumption value of the daily resource to obtain the initial total fatigue, and then compares it with the basic fatigue threshold. If it exceeds, the adjustment process is triggered. The adjustment strategy uses a greedy algorithm, which removes the resource with the highest fatigue consumption value each time until the threshold constraint is met. This iterative optimization mechanism ensures that the initial resource combination not only meets the time constraints, but also satisfies the fatigue limit, laying the foundation for subsequent balanced optimization.

[0096] In the embodiment of the application, the detailed implementation steps for obtaining the actual fatigue distribution of the daily trip according to the resource fatigue consumption value of the daily travel resource in the initial travel resource combination and the dynamic fatigue adjustment coefficient include:

[0097] The resource fatigue consumption value of the daily travel resource in the initial travel resource combination is time-weighted, wherein the weight of the time weighting is determined by the tour order of the travel resource in the daily trip, and the weight is higher when the tour order is later. Time weighting is a key technology that considers the cumulative fatigue effect, which reflects the difference in the perception of the same fatigue consumption at different times. The weighting process uses an exponential growth model, and the amplification effect of fatigue perception in the later period is more obvious. The weighting formula is:

[0098] ; wherein, is the weighted fatigue of the th resource on the th day, is the original fatigue consumption value of the resource, is the time weighting coefficient (usually 0.5-1.0), is the sequence number of the resource in the itinerary of the day, is the total number of resources in the itinerary of the day. is the total number of resources in the itinerary of the day.

[0099] The weighting formula reflects the nonlinear characteristics of fatigue accumulation. Activities of the same intensity will feel more tired at the end of the day than at the beginning, which is consistent with human physiology. Time weighting makes fatigue assessment more consistent with actual experience, providing accurate basis for subsequent optimization.

[0100] The product of the time-weighted resource fatigue consumption value and the dynamic fatigue adjustment coefficient is recorded as the weighted fatigue of the daily itinerary. Weighted fatigue is the final fatigue index that considers both timing factors and physical state, directly reflecting the actual experience of tourists. The calculation formula is:

[0101] ; wherein, is the weighted fatigue of the day, is the dynamic fatigue adjustment coefficient of the day, is the dynamic fatigue adjustment coefficient of the day, is the sum of the time-weighted fatigue of all resources in the day. The calculation logic of weighted fatigue considers two key factors: first, the fatigue consumption of the resource itself and its timing effect; second, the influence of the physical state of tourists on the day on fatigue perception. The dynamic adjustment coefficient, as a multiplicative factor, amplifies or reduces the overall fatigue, achieving personalized fatigue assessment.

[0102] According to the weighted fatigue of the daily itinerary, the fatigue distribution curve within the total number of days of the itinerary is constructed. The fatigue distribution curve is a visual expression of the overall fatigue state, which intuitively shows the change law of fatigue on each day. The construction process takes the daily weighted fatigue as discrete points, generates a continuous and smooth curve through cubic spline interpolation, ensuring the continuity and smoothness of the curve. The distribution curve not only reflects the absolute fatigue level, but also embodies the trend of fatigue, providing an intuitive basis for fatigue balance optimization.

[0103] According to the weighted fatigue of the daily itinerary, the fatigue distribution curve within the total number of days of the itinerary is constructed. The fatigue distribution curve is a visual expression of the overall fatigue state, which intuitively shows the change law of fatigue on each day. The construction process takes the daily weighted fatigue as discrete points, generates a continuous and smooth curve through cubic spline interpolation, ensuring the continuity and smoothness of the curve. The distribution curve not only reflects the absolute fatigue level, but also embodies the trend of fatigue, providing an intuitive basis for fatigue balance optimization.

[0104] ​The fatigue degree distribution curve is smoothed to obtain an actual fatigue degree distribution. The smoothing is an important step of eliminating random fluctuations and highlighting the main trend, so that the fatigue degree evaluation is more stable and reliable. The processing process adopts a Gaussian smoothing algorithm, eliminates local abnormal fluctuations through weighted average, and retains the overall change trend. The smoothing parameter is dynamically adjusted according to the number of days, the more the days, the larger the smoothing window, to ensure the representativeness and reliability of the results. The smoothed fatigue degree distribution is closer to the actual experience, providing an accurate reference for subsequent balanced optimization.

[0105] In the embodiment of the application, based on the actual fatigue degree distribution and the basic fatigue degree threshold, the initial travel resource combination is dynamically adjusted to obtain the detailed implementation steps of the optimized travel resource combination with balanced fatigue degree.

[0106] The difference between the weighted fatigue degree of each daily itinerary in the actual fatigue degree distribution and the basic fatigue degree threshold is calculated, denoted as the daily fatigue degree over-standard value. The fatigue degree over-standard value is a direct basis for judging whether adjustment is needed, and quantifies the degree of exceeding the bearing capacity of tourists. The calculation formula is:

[0107] ; wherein, is the fatigue degree over-standard value of the th day, is the weighted fatigue degree of the th day, is the basic fatigue degree threshold of the th day.

[0108] The over-standard value greater than zero indicates that the fatigue degree of the day exceeds the threshold and needs to be adjusted; the greater the over-standard value, the higher the priority of adjustment. This index directly guides the order and intensity of subsequent resource adjustment, ensuring the effectiveness and pertinence of the optimization process.

[0109] If the daily fatigue degree over-standard value is greater than zero, in the initial travel resource combination of the corresponding day, identify the travel resource with the highest resource fatigue degree consumption value, denoted as the to-be-adjusted travel resource. This step determines the specific adjustment object, adopts the "highest fatigue degree priority" strategy, and adjusts the resource with the greatest impact on fatigue degree first to ensure the most significant adjustment effect. The identification process considers the comprehensive influence of the original fatigue degree consumption value and the time sequence position, and quickly locates the target resource through a sorting algorithm. The to-be-adjusted resource is usually 1-2 resources with the greatest fatigue degree contribution in the daily itinerary, and targeted adjustment can significantly improve the overall fatigue degree distribution.

[0110] In the candidate travel resource set, search for a substitute travel resource whose interest matching degree with the to-be-adjusted travel resource is greater than a preset matching degree threshold and whose resource fatigue consumption value is lower than that of the to-be-adjusted travel resource. The substitute resource search is the core step of fatigue optimization, and the goal is to find a substitute option with lower fatigue under the premise of maintaining interest matching. The search process adopts multi-constraint condition screening: the interest matching degree needs to be higher than the preset threshold (usually 0.75), to ensure that the substitution does not reduce the interest satisfaction degree; the fatigue consumption value needs to be lower than the original resource, to ensure that the adjustment effectively reduces fatigue; and the tour duration needs to be suitable for the remaining time of the day, to ensure that the time arrangement is reasonable. This multi-dimensional screening ensures that the found substitute resource meets the fatigue optimization goal and does not sacrifice the tourist experience.

[0111] According to the resource tour duration and resource geographic location of the substitute travel resource, it is judged whether it meets the time window constraint of the corresponding day's time sequence dependency graph; if it meets, the to-be-adjusted travel resource is replaced with the substitute travel resource, and the initial travel resource combination is updated. The time window constraint verification is a key step to ensure the feasibility of adjustment, to avoid producing an unexecutable itinerary arrangement. The verification process is based on the time sequence dependency graph model, calculates the total tour duration and path time consumption after the substitute resource is integrated into the daily itinerary, and compares it with the daily maximum tour duration. If the constraint is met, the replacement operation is performed, and the resource combination and related parameters are updated; if it is not met, the next candidate substitute resource is tried until a substitute scheme that meets the condition is found or all possibilities are exhausted. This constraint verification mechanism ensures the practicality and executability of the optimization process.

[0112] Repeat the above steps until the weighted fatigue degree of all days in the actual fatigue degree distribution is less than or equal to the basic fatigue threshold, and obtain the optimized travel resource combination. Iterative optimization is a key mechanism to achieve overall fatigue balance, and the best state is gradually reached through multiple rounds of adjustment. The optimization process adopts a "greedy + backtracking" strategy, and each round prioritizes processing the most serious over-standard day. If new over-standard situations occur after adjustment, backtrack and re-plan. The iteration termination condition is that the overall fatigue degree is not over-standard, or the maximum iteration number (usually 3 times the number of itinerary days) is reached. The final optimized travel resource combination not only meets the fatigue constraint, but also maintains a high degree of interest matching, providing a high-quality resource basis for path planning.

[0113] In the embodiment of the application, the detailed implementation steps of obtaining the path planning result of the daily itinerary according to the resource geographic location and resource tour duration of each daily travel resource in the optimized travel resource combination include:

[0114] The resource geographic locations of each daily travel resource in the optimized travel resource combination are spatially clustered to obtain the geographic partitions of each daily itinerary. Spatial clustering is a preprocessing step for optimizing the tour order, grouping resources with similar geographic locations, and reducing unnecessary back-and-forth travel. The clustering process uses the DBSCAN (Density-Based Spatial Clustering) algorithm to automatically determine the number and range of partitions based on the geographic distance between resources, without the need to pre-set the number of partitions. Key parameters include the neighborhood radius (usually set to 2-5 kilometers, adjusted according to city size) and the minimum number of points (usually set to 2). The clustering results form several geographic partitions, with resources in each partition having relatively concentrated geographic locations, facilitating efficient touring. Geographic partitions not only optimize the complexity of path planning, but also better align with actual touring habits, as tourists often prefer to visit multiple attractions within the same area.

[0115] Within each geographic partition, the optimal tour order is constructed based on resource touring duration and path time consumption between travel resources, where the optimal tour order is determined using a dynamic programming-based Traveling Salesman Problem (TSP) solving algorithm. TSP is a classic combinatorial optimization problem, aiming to find a closed-loop path that visits all points once with the shortest total path. In this system, the problem is transformed into finding a sequence that visits all resources in a partition with the shortest total path time consumption under time constraints. The solution uses an improved dynamic programming algorithm, which improves computational efficiency through state compression techniques, reducing complexity from O(n!) to O(n²×2 n ), where n is the number of resources in the partition. The algorithm design takes into account time window constraints (such as attraction opening hours) and touring duration, ensuring that the generated path is both efficient and feasible. Determining the optimal tour order within a partition significantly reduces the time spent on inefficient movement during the trip, improving touring efficiency.

[0116] Based on the spatial adjacency relationship of geographic partitions, the partition tour order of each daily itinerary is obtained. The partition tour order is a macro-level path planning that determines the order of visiting each partition. The acquisition process is based on the spatial adjacency graph between partitions, treating partitions as nodes and the connectivity between partitions as edges to construct an undirected graph model. The partition tour order is determined by an improved nearest neighbor algorithm, which starts from the accommodation or starting point and selects the nearest unvisited partition each time until all partitions are covered. The algorithm takes into account the traffic convenience and time cost between partitions to ensure the efficiency of transition between partitions. The optimization of the partition tour order avoids repeated back-and-forth travel across regions, making the overall itinerary more coherent and reasonable.

[0117] Based on the partition tour order and the optimal tour order within the partition, a path planning result is generated. The path planning result is a detailed daily tour plan, including specific visiting order, estimated time and transfer mode. The generation process combines macro partition order and micro resource order to form a complete tour route. The planning result considers the path from the starting point (such as a hotel) to the first attraction and the path from the last attraction back to the end point, forming a complete closed loop. At the same time, the path time estimation is dynamically adjusted according to the traffic conditions at different times, improving the accuracy of the planning. The final path planning result optimizes the overall tour efficiency and ensures the best tour experience of each resource within the partition, providing detailed and feasible route guidance for trip recommendation.

[0118] In the embodiment of the present application, the detailed implementation steps for obtaining the geographic accessibility score of each initial candidate cultural and tourism resource based on the resource geographic location of the initial candidate cultural and tourism resource and the trip starting location in the trip planning demand information include:

[0119] The straight-line distance between the resource geographic location of the initial candidate cultural and tourism resource and the trip starting location is calculated, denoted as the initial distance. The initial distance is a basic index for geographic accessibility evaluation, reflecting the absolute geographic location advantage of the resource. The calculation uses the formula in the spherical coordinate system, considering the influence of the curvature of the earth to ensure the accuracy of the distance calculation. The formula is:

[0120] ; wherein, is the radius of the earth, , is the latitude and longitude of the starting location, , is the latitude and longitude of the resource location, is the initial distance.

[0121] The unit of the initial distance is kilometers, and the smaller the value, the closer the location and the higher the basic accessibility. This index directly reflects the basic distance cost to reach the resource and is the primary factor in accessibility evaluation.

[0122] Based on the resource geographic location of the initial candidate cultural and tourism resource, the traffic convenience of the area where it is located is obtained, wherein the traffic convenience is determined by the density of public transportation stations and road density around the resource geographic location. Traffic convenience is a key environmental factor for evaluating accessibility, reflecting the degree of perfection of regional transportation infrastructure. The acquisition process is based on geographic information systems, setting a buffer zone (usually 1-2 kilometers) around the resource location, counting the number of public transportation stations and the total length of road network in the area, and calculating the density value per unit area. The formula for calculating traffic convenience is:

[0123] ; wherein, is the traffic convenience, the number of public transportation stations, the total length of road network, the area of buffer zone, and the weight coefficient, and .

[0124] The weight setting is determined according to the traffic preference of tourists. The tourists who prefer public transportation have a higher value, and the tourists who prefer self-driving have a higher value. The traffic convenience is a dimensionless index, usually ranging from [0, 1], and the higher the value, the more convenient the traffic and the easier to reach.

[0125] The initial distance is negatively correlated to obtain the distance score. The negative correlation mapping is a key step to convert the distance into a score, which realizes the intuitive logic of "the farther the distance, the lower the score". The mapping adopts an exponential decay function, and the calculation formula is:

[0126] ; wherein, is the distance score, is the initial distance, is the distance decay parameter (usually set to 1 / 3 of the city radius or the travel coverage range).

[0127] When the distance is 0, the score is 1; as the distance increases, the score gradually decays; when the distance is much larger than , the score approaches 0. The distance score takes a value in the range (0, 1], which intuitively reflects the level of accessibility based on distance.

[0128] The distance score and the weighted sum of traffic convenience are normalized to obtain the geographic accessibility score, wherein the weights of the weighted sum are determined by the traffic tool preference of tourists. The geographic accessibility score is the final evaluation index considering distance and traffic conditions comprehensively, which fully reflects the difficulty of reaching the resource. The calculation formula is:

[0129] ; wherein, is the geographic accessibility score, is the normalization coefficient, is the distance score, is the traffic convenience, and is the weight coefficient, and .

[0130] The weight setting is determined according to the traffic characteristics of tourists: for self-driving tourists, the distance factor is more important, and for public transportation tourists, The value is higher, and the traffic convenience is more critical. The normalization processing ensures that the final score is distributed in the [0, 1] interval, facilitating subsequent screening and comparison. The higher the geographical accessibility score, the easier the resource is to reach, and the more suitable it is to be included in the trip planning.

[0131] In the embodiment of the application, the fatigue distribution curve is smoothed, comprising:

[0132] A Gaussian kernel function is constructed, wherein the standard deviation of the Gaussian kernel function is determined by the total number of days of the trip and the daily physical fluctuation amplitude of the physical preference curve. The Gaussian kernel function is a classic tool for signal smoothing, which eliminates random fluctuations through weighted averaging and retains the main trend. The function construction adopts a standard Gaussian distribution form, and the key parameter is the standard deviation , which determines the smoothing degree. The standard deviation calculation formula is:

[0133] ; wherein, is the standard deviation of the Gaussian kernel function, 1 is an adjustment coefficient (usually 0.3-0.5), is the total number of days of the trip, is the average daily physical fluctuation amplitude of the physical preference curve.

[0134] The longer the trip, the larger the smoothing window should be to avoid excessive sensitivity; the greater the physical fluctuation, the lower the smoothing degree should be to retain the necessary change information. The constructed Gaussian kernel function can effectively eliminate noise and retain the true trend of the fatigue distribution, providing an ideal tool for smoothing.

[0135] The fatigue distribution curve is discretized into a daily fatigue sequence. Discretization is a preprocessing step for convolution operation, which converts continuous curves into discrete numerical sequences for subsequent calculation. The processing process adopts equal time interval sampling, and the number of sampling points is determined according to the total number of days of the trip, and each sampling point corresponds to a day's fatigue value. The discretization result is a numerical sequence with a length of (total number of days of the trip), which accurately expresses the key characteristics of the original fatigue distribution and provides input data for convolution operation.

[0136] The daily fatigue sequence and the Gaussian kernel function are convolved to obtain the smoothed fatigue sequence. Convolution operation is the core step to achieve smoothing effect, which eliminates local fluctuations through weighted averaging and highlights overall trend. The calculation adopts the standard discrete convolution formula:

[0137] ; wherein, is the smoothed fatigue of the day, is the weight of the th Gaussian kernel function, is the original th The daily fatigue degree is summed in the range [-m, m], and m is the radius of the convolution window (usually taken as The corresponding integer value is obtained.

[0138] The convolution operation considers the boundary processing problem, and uses the mirror padding method to process the two ends of the sequence to ensure the continuity and reliability of the smooth result. The smoothed fatigue degree sequence retains the main trend of the original distribution, while effectively eliminating the influence of random fluctuations and outliers, providing a stable and reliable reference for subsequent optimization.

[0139] The smoothed fatigue degree sequence is interpolated to restore a continuous curve, denoted as the actual fatigue degree distribution. Interpolation restoration is a key step to convert discrete sequences into continuous functions, which facilitates visual display and accurate calculation. The restoration process uses cubic spline interpolation method to ensure the continuity and smoothness of the curve, avoiding the polyline effect caused by linear interpolation. The selection of interpolation points is based on the original sampling points, while increasing the interpolation density in the key change area to ensure that the restored curve can accurately reflect the trend of fatigue degree. The final actual fatigue degree distribution is a smooth and continuous curve, which intuitively shows the dynamic changes of fatigue degree during the trip, providing an intuitive and reliable evaluation basis for fatigue degree balance optimization.

[0140] In the embodiment of the present application, the detailed implementation steps of outputting the final trip recommendation scheme based on the path planning result and the interest preference vector include:

[0141] According to the path planning result, a time sequence tour plan of daily trip is generated, wherein the time sequence tour plan includes the tour order of daily travel resources, the expected tour duration and the path time consumption. The time sequence tour plan is a specific execution guide of the trip, which intuitively shows the detailed arrangement of each day. The generation process is based on the path planning result, combined with the opening time and recommended tour duration of each resource, to determine the specific access time period of each resource. The plan is displayed in the form of time axis, clearly marking the start time, end time and path time consumption of each link, facilitating the tourists to grasp the overall rhythm. At the same time, appropriate rest time and meal time are added in the plan, considering the physiological needs in the actual tourism process, to ensure the rationality and executability of the trip. The time sequence tour plan is both detailed and flexible, providing clear trip guidance for tourists while leaving appropriate space for free adjustment.

[0142] According to the interest preference vector, the interest matching degree of each cultural and travel resource in the optimized cultural and travel resource combination is obtained. The interest matching degree is the core basis for resource recommendation, and directly shows the degree of matching between the resource and the interest of the tourists. The acquisition process is based on cosine similarity calculation, and the feature vector of each resource is compared with the interest preference vector of the tourists to obtain a matching score between 0 and 1. The matching degree calculation not only considers the overall similarity, but also pays special attention to the specific dimension preferred by the tourists, adopts a weighted cosine similarity method, and enhances the influence of the key dimension. The calculation result forms the interest matching degree score of each resource, and ranks the resources to identify the core resources and characteristic resources that best match the interests of the tourists, and provides data support for the generation of recommendation reasons.

[0143] The interest matching degree is associated with the time sequence tour plan to generate the recommendation reasons for each daily itinerary, wherein the recommendation reasons include the interest matching degree ranking of the cultural and travel resources and the characteristic description of the resources. The recommendation reasons are the value explanation of the itinerary scheme, which helps the tourists to understand the basis and highlights of the recommendation. The generation process first identifies the resources with high matching degree ranking in the daily itinerary (usually the top 30%), and marks them as the highlights of the day; then, the personalized recommendation language is generated by combining the characteristic labels of the resources. The recommendation reasons adopt natural language generation technology, dynamically adjust the expression style according to the resource type and the interest of the tourists, emphasize the cultural value for historical and cultural resources, highlight the visual experience for natural scenery resources, and focus on the interactive fun for experience activities resources. This personalized recommendation reason not only explains why the system recommends these resources, but also stimulates the tourists' interest and expectation in the tour, and enhances the attractiveness and persuasiveness of the itinerary scheme.

[0144] The time sequence tour plan and the recommendation reasons are integrated into a visual itinerary recommendation scheme, which is output to the user terminal. The visual itinerary recommendation scheme is the final output of the system, which directly shows the complete tourism planning. The integration process adopts a multi-level information architecture design, the top layer is the itinerary overview (total days, covered resources, highlight recommendation), the middle layer is the daily itinerary arrangement (time table, route map, resource list), and the bottom layer is the resource details (introduction, picture, recommendation reason). The visual display adopts an interactive map combined with a time axis, supports zooming and detail browsing, and provides an immersive planning experience. At the same time, the scheme supports multi-terminal adaptation (PC, mobile device, printed version), which meets the use requirements in different scenarios. The final itinerary recommendation scheme is professional and friendly, has rigorous time planning and warm personalized recommendation, and fully meets the needs of tourists for high-quality tourism experience.

[0145] The present application realizes multi-dimensional intelligent recommendation and dynamic combination of cultural and travel resources through interest preference analysis, physical state evaluation, fatigue balance optimization and path intelligent planning. The adaptive optimization characteristics of the present application can adjust the recommendation parameters in real time according to the individual differences of the tourists and the planning requirements, significantly improve the tourism experience, and reduce the fatigue of the tourists.

[0146] The above merely provides preferred embodiments of the present application but not for limiting the present application. Although the present application has been described in detail with reference to the foregoing embodiments, the technical solutions recorded in the foregoing embodiments can still be modified or some technical features can be replaced by equivalents for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0147] It should be noted that the formula in the specification is a dimensionless value, and the formula is obtained by software simulation of a large amount of data to obtain the most real situation. The preset parameters and threshold values in the formula are set by those skilled in the art according to the actual situation.

[0148] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirit of the present application. The scope of the present application is defined by the claims and their equivalents.

Claims

1. A multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources, characterized in that: include: The user demand acquisition module is used to acquire tourists' itinerary planning demand information, which includes the total number of days of the trip, the maximum daily sightseeing time, interest preference vector, and physical fitness preference curve. The physical fitness parameter calculation module is used to segment the physical fitness preference curve and obtain the physical fitness preference value for each day within the total number of days of the trip; The basic fatigue threshold for each day is obtained by multiplying the physical fitness preference value by the preset fatigue baseline value. The trend of the physical fitness preference curve over the total number of days of the trip was statistically analyzed to obtain the daily physical fitness decline slope and daily physical fitness fluctuation range of the physical fitness preference curve. The daily baseline adjustment factor is obtained based on the ratio of the daily physical strength decline slope to a preset slope threshold. The daily fluctuation adjustment factor is obtained based on the ratio of the daily physical strength fluctuation range to the preset fluctuation threshold. The weighted sum of the basic adjustment factor and the fluctuation adjustment factor is normalized to obtain the dynamic fatigue adjustment coefficient. The resource matching and filtering module is used to obtain a set of candidate cultural and tourism resources that match tourists' interests and preferences based on the interest preference vector and the cultural and tourism resource feature vector in the cultural and tourism resource database. The initial resource combination module is used to obtain the initial cultural and tourism resource combination for each day within the total number of days of the trip, based on the resource fatigue consumption value, resource tour duration and resource geographical location of each cultural and tourism resource in the candidate cultural and tourism resource set, combined with the maximum daily tour duration and the basic fatigue threshold for each day. The fatigue balance optimization module is used to perform time-series weighting on the daily fatigue consumption values ​​of cultural and tourism resources in the initial cultural and tourism resource combination. The weight of the time-series weighting is determined by the order of visiting cultural and tourism resources in the daily itinerary. The later the visit order, the higher the weight. The product of the time-weighted resource fatigue consumption value and the dynamic fatigue adjustment coefficient is recorded as the weighted fatigue of the daily schedule. Based on the weighted fatigue level of the daily itinerary, construct a fatigue level distribution curve over the total number of days of the itinerary; The fatigue distribution curve is smoothed to obtain the actual fatigue distribution. The difference between the weighted fatigue level of each day's journey in the actual fatigue level distribution and the basic fatigue level threshold is calculated and recorded as the daily fatigue level exceeding the standard value. If the daily fatigue level exceeds zero, then in the initial cultural and tourism resource combination for the corresponding day, identify the cultural and tourism resource with the highest fatigue level consumption value and record it as the cultural and tourism resource to be adjusted. In the candidate set of cultural and tourism resources, search for alternative cultural and tourism resources that have an interest matching degree greater than a preset matching degree threshold and a resource fatigue consumption value lower than that of the cultural and tourism resources to be adjusted. Based on the visit duration and geographical location of the alternative cultural and tourism resources, determine whether they meet the time window constraints of the time series dependency graph for the corresponding day; if they do, replace the cultural and tourism resources to be adjusted with the alternative cultural and tourism resources and update the initial combination of cultural and tourism resources. Repeat the above steps until the weighted fatigue degree of all days in the actual fatigue degree distribution is less than or equal to the basic fatigue degree threshold, to obtain the optimized combination of cultural and tourism resources; The route planning and recommendation output module is used to obtain the route planning results for each day's itinerary based on the geographical location and visit duration of the daily cultural and tourism resources in the optimized cultural and tourism resource combination; and to output the final itinerary recommendation scheme based on the route planning results and the interest preference vector.

2. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 1, characterized in that, The step of obtaining a set of candidate cultural and tourism resources that match tourists' interests and preferences based on the interest preference vector and the cultural and tourism resource feature vector in the cultural and tourism resource database includes: Cluster analysis is performed on the resource feature vector of each cultural and tourism resource in the cultural and tourism resource database to obtain multiple cultural and tourism resource category clusters; Calculate the Euclidean distance between the interest preference vector and the cluster center of each cultural and tourism resource category cluster, and obtain the top K cultural and tourism resource category clusters with the highest matching degree with the interest preference vector, where K is a preset threshold for the number of category clusters; In the first K cultural and tourism resource category clusters, the cosine similarity between the resource feature vector and the interest preference vector of each cultural and tourism resource is calculated, and cultural and tourism resources with a cosine similarity greater than a preset similarity threshold are recorded as initial candidate cultural and tourism resources. Based on the geographical location of the initial candidate cultural and tourism resources and the starting location of the trip in the trip planning requirements information, the geographical accessibility score of each initial candidate cultural and tourism resource is obtained; the initial candidate cultural and tourism resources are filtered based on the geographical accessibility score to obtain the set of candidate cultural and tourism resources.

3. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 1, characterized in that, The step of obtaining the initial combination of cultural and tourism resources for each day within the total number of days of the trip, based on the resource fatigue consumption value, visit duration, and geographical location of each cultural and tourism resource in the candidate cultural and tourism resource set, combined with the maximum daily visit duration and the basic daily fatigue threshold, includes: Based on the visit duration and geographical location of each cultural and tourism resource in the candidate cultural and tourism resource set, calculate the path time between each cultural and tourism resource; Based on the path time and the resource tour duration, a temporal dependency graph is constructed between cultural and tourism resources, wherein the nodes of the temporal dependency graph are cultural and tourism resources, and the edges are the path time between cultural and tourism resources; In the time-series dependency graph, based on the maximum daily tour duration, a graph partitioning algorithm with time window constraints is used to divide the candidate cultural and tourism resource set into daily initial subsets within the total number of days of the trip; For each cultural and tourism resource in the daily initial subset, calculate the sum of its resource fatigue consumption values, and record it as the daily initial fatigue value. Compare the daily initial fatigue value with the daily basic fatigue threshold. If the daily initial fatigue value is greater than the basic fatigue threshold, remove the cultural and tourism resource with the highest resource fatigue consumption value from the daily initial subset, and recalculate the daily initial fatigue value until the daily initial fatigue value is less than or equal to the basic fatigue threshold to obtain the initial combination of cultural and tourism resources.

4. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 1, characterized in that, The step of obtaining the daily itinerary route planning results based on the geographical location and visit duration of the daily cultural and tourism resources in the optimized cultural and tourism resource combination includes: Spatial clustering is performed on the geographical locations of daily cultural and tourism resources in the optimized cultural and tourism resource combination to obtain the geographical partitions of daily itineraries; Within each geographical region, the optimal tour order is constructed based on the tour duration of the resources and the path time between cultural and tourism resources. Based on the spatial adjacency of the geographical partitions, the daily itinerary's partition tour order is obtained; based on the partition tour order and the optimal tour order within the partition, the route planning result is generated.

5. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 2, characterized in that, The step of obtaining the geographical accessibility score for each initial candidate cultural and tourism resource based on its geographical location and the starting location of the itinerary in the itinerary planning requirements includes: Calculate the straight-line distance between the geographical location of the initial candidate cultural and tourism resources and the starting location of the itinerary, and record it as the initial distance; Based on the geographical location of the initial candidate cultural and tourism resources, the transportation convenience of the area where they are located is obtained, wherein the transportation convenience is determined by the density of public transportation stations and roads around the geographical location of the resources; The initial distance is negatively correlated to obtain a distance score; the weighted sum of the distance score and the transportation convenience is normalized to obtain the geographical accessibility score, wherein the weights of the weighted sum are determined by the tourist's transportation preferences.

6. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 1, characterized in that, The smoothing process for the fatigue distribution curve includes: Construct a Gaussian kernel function, wherein the standard deviation of the Gaussian kernel function is determined by the total number of days of the trip and the daily physical fitness fluctuation range of the physical fitness preference curve; The fatigue distribution curve is discretized into a daily fatigue sequence; The daily fatigue rate sequence is convolved with the Gaussian kernel function to obtain a smoothed fatigue rate sequence. The smoothed fatigue degree sequence is interpolated to restore it to a continuous curve, which is denoted as the actual fatigue degree distribution.

7. The multi-dimensional intelligent recommendation and dynamic combination system for cultural and tourism resources according to claim 1, characterized in that, The final itinerary recommendation scheme is output based on the path planning results and the interest preference vector, including: Based on the route planning results, a time-series tour plan for each day's itinerary is generated, wherein the time-series tour plan includes the order of visiting cultural and tourism resources each day, the estimated tour duration, and the route time. Based on the interest preference vector, obtain the interest matching degree of each cultural and tourism resource in the optimized cultural and tourism resource combination; The interest matching degree is associated with the time-series tour plan to generate a recommendation reason for each day's itinerary, wherein the recommendation reason includes the interest matching degree ranking of cultural and tourism resources and the description of the resource features; The time-series tour plan and the reasons for recommendation are integrated into a visual itinerary recommendation scheme and output to the user terminal.

Images