Travel itinerary planning method based on artificial intelligence

Abstract

The invention discloses a travel itinerary planning method based on artificial intelligence, and belongs to the technical field of itinerary planning, and the method specifically comprises the steps: obtaining a historical itinerary sequence of a group and corresponding itinerary evaluation data, and constructing an itinerary-evaluation mapping library; identifying a specific scenic spot based on the mapping library, and extracting a cosexual behavior offset feature value after the user accesses the scenic spot; clustering is carried out according to the feature values to obtain behavior feature clusters, and subsequent scenic spots which bring higher experience feedback to the behavior feature clusters are screened out through statistical test to form an association rule base; generating an initial itinerary plan according to a current user request and the rule base; and in the travel execution process, the offset category of the travel is judged according to user behavior data collected in real time, the rule base is inquired according to the category to obtain an experience optimization decision set, and the remaining travel is dynamically adjusted. According to the invention, the transformation from static resource matching to dynamic experience management is realized, and the individuation and satisfaction of travel itinerary planning are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of itinerary planning technology, and more specifically to a tourism itinerary planning method based on artificial intelligence. Background Technology

[0002] In the digital tourism sector, AI-based itinerary planning services have become an important tool for improving travel efficiency and personalized experiences for tourists. Existing intelligent itinerary planning methods typically combine weather conditions with attraction popularity, explicit user preferences, geographical proximity, and historical access data to achieve static optimization through multi-dimensional environmental adaptation. By incorporating meteorological factors, direct obstacles and explicit risks caused by weather can be mitigated to some extent. For example, recommending outdoor sightseeing on sunny days or guiding users to indoor venues on rainy or snowy days can improve the basic feasibility and safety of the itinerary. While satisfying user interests and geographical efficiency, it also considers the environmental conditions of the day, making the recommendations more comprehensive across multiple static dimensions and improving the user experience when making real-time activity choices under specific weather conditions.

[0003] However, these mainstream methods generally suffer from a significant technical limitation: their planning logic is essentially static and based on resource tag matching, severely neglecting the dynamic nature of tourism as a high-intensity physical and cognitive load experience. Specifically, existing technical solutions lack the ability to model and address the key factor of "cumulative physiological fatigue during sightseeing" and the resulting "decline in the quality of subsequent experiences." They cannot predict or perceive how the significant changes in a user's physiological state (such as fatigue level) after completing a high-intensity visit (e.g., a mountainous area requiring long hikes) will fundamentally affect their satisfaction with subsequent activities.

[0004] Current solutions largely remain at the level of post-trip feedback or simple filtering. For example, some allow users to manually set the pace of the trip (such as "easy" or "intense"), but this is a static, subjective, pre-set configuration that cannot adapt to the real-time changes in physical condition during the trip. Other solutions monitor indicators such as heart rate using wearable devices, but their applications are mostly limited to health alerts or exercise analysis, without deep intelligent coupling and dynamic reorganization with the trip content (such as the type and intensity of the next attraction). More importantly, existing technologies lack the ability to utilize massive amounts of anonymized historical data to scientifically analyze and verify "which subsequent activities yield a better experience under what fatigue states." They cannot answer a core question: when many tourists experience leg soreness after visiting Huangshan, does historical data indicate whether they ultimately achieved higher overall satisfaction by visiting Hongcun Ancient Town or an indoor museum the following day?

[0005] Therefore, existing technologies lead to a common user experience dilemma: a route might be recommended that appears optimal based on tourist attraction popularity and geographical logic, but in actual use, user fatigue leads to a decline in experience throughout the trip, creating the contradiction of "the plan seems perfect, but the experience is actually poor." This stems from the fact that existing technologies, in their technical architecture, have severed the closed-loop data connection between "trip planning," "behavioral status perception," and "experience result feedback." Summary of the Invention

[0006] The purpose of this invention is to provide an artificial intelligence-based method for planning travel itineraries, addressing the following technical problems: Existing intelligent itinerary planning methods based on static interest matching and popularity ranking cannot dynamically adapt to the real-time physiological fatigue state of users during the trip, resulting in a mismatch between subsequent recommendations and the user's actual experience, and causing a decline in the overall quality of the travel experience.

[0007] The objective of this invention can be achieved through the following technical solutions: The AI-based travel itinerary planning method includes the following steps: S1. Obtain the historical travel sequences and corresponding travel evaluation data of anonymous user groups in the target area, and construct a travel-evaluation mapping library; S2, perform group behavior analysis on the itinerary-evaluation mapping library, identify common behavior shift feature values ​​after visiting a specific attraction, and cluster the attraction visit records in the historical itinerary sequence that are located after the specific attraction based on the common behavior shift feature values ​​to form an association rule library between behavior shift feature values ​​and subsequent attraction selection clusters. S3, responding to the current user's planning request, and generating an initial itinerary plan containing a sequence of attraction visits based on the association rule base; S4, during the process of the current user executing the initial itinerary planning, the user's location sequence and dwell time sequence are obtained in real time and the behavior feature vector is calculated. The behavior feature vector is input into the preset behavior offset classification model and the offset category to which the current user's real-time behavior belongs is output. S5. Obtain the corresponding target attraction selection cluster from the association rule base according to the offset category, dynamically adjust the remaining itinerary plan of the current user based on the target attraction selection cluster, and synchronize it to the current user's terminal device.

[0008] As a further aspect of the present invention: the specific construction process of the trip-evaluation mapping library in S1 is as follows: Obtain the historical itinerary records and historical evaluation records of anonymous user groups; parse out the attraction visit sequence indexed by itinerary identifiers from the historical itinerary records, the attraction visit sequence containing attraction identifiers arranged by timestamps; The evaluation data corresponding to each itinerary identifier is parsed from the historical evaluation record set. The evaluation data includes numerical overall score data and text-based comment data. Sentiment analysis is performed on the text-based comment data to extract the sentiment tags and intensity values ​​that represent experience tendencies and generate sentiment vectors. The numerical overall rating data is concatenated with the emotion vector to form an experience feedback vector corresponding to the itinerary identifier; the itinerary identifier is used as the association key to bind the attraction visit sequence with the experience feedback vector corresponding to the itinerary identifier, generating an itinerary-evaluation mapping record; all itinerary-evaluation mapping records corresponding to all itinerary identifiers are collected to generate an itinerary-evaluation mapping library.

[0009] As a further aspect of the present invention: in step S2, the specific process for acquiring a particular scenic spot is as follows: Extract all attraction visit sequences from the trip-evaluation mapping database, and aggregate the attraction identifiers to form an initial attraction set; for each attraction identifier in the initial attraction set, retrieve all trip-evaluation mapping records containing that identifier from the trip-evaluation mapping database to form an associated record set; For each record in the associated record set, locate the visit time point of the attraction identifier in its attraction visit sequence, divide the sequence into a preceding subsequence and a subsequent subsequence based on the time point; obtain the mean local experience feedback corresponding to the preceding subsequence and the subsequent subsequence respectively, and calculate the difference between the two as the individual experience feedback attenuation degree of the record. Statistical analysis is performed on the individual experience feedback decay rate of all records in the associated record set to obtain the group experience feedback decay rate and its statistical distribution parameters of the scenic spot identifier; after traversing the initial scenic spot set, the group experience feedback decay rate and its statistical distribution parameters of each scenic spot identifier are filtered through statistical significance test. Scenic spots with negative group experience feedback decay and statistical significance exceeding a preset threshold are marked as specific scenic spots, forming a specific scenic spot set.

[0010] As a further aspect of the present invention: the specific process of the common behavior offset feature value in S2 is as follows: For each specific attraction in the specific attraction set, extract its corresponding associated record set from the itinerary-evaluation mapping library; filter out records from the associated record set whose individual experience feedback decay is negative and whose absolute value exceeds a first preset threshold to form a high decay record subset; For each record in the high-attenuation record subset, the trajectory data within a fixed time window starting from the visit time of the specific scenic spot in its scenic spot visit sequence is extracted; based on the trajectory data, the rate of change of movement speed and the rate of change of dwell event frequency of the record within the fixed time window compared to a control time window before the visit time point are calculated; The rate of change of movement speed and the rate of change of dwell event frequency of all records in the high-attenuation record subset are collected, and their arithmetic mean and standard deviation are calculated respectively. The combination of the arithmetic mean and standard deviation of the rate of change of movement speed and the arithmetic mean and standard deviation of the rate of change of dwell event frequency is used as the common behavior offset feature value corresponding to the specific scenic spot.

[0011] As a further aspect of the present invention: in S2, the specific construction process of the association rule base is as follows: For each specific attraction in the set of specific attractions, obtain its high-attenuation record subset; perform cluster analysis based on the common behavioral offset feature values ​​of each record in the high-attenuation record subset to form multiple behavioral feature clusters; For each record in the high-decay record subset, extract the identifier of the first subsequent attraction following that specific attraction in its attraction visit sequence and its corresponding experience feedback vector; The subsequent attraction identifiers and the experience feedback vectors are assigned to the corresponding behavioral feature clusters; for each behavioral feature cluster, the records within it are grouped according to the subsequent attraction identifiers to form multiple candidate option subsets; Calculate the mean of the experience feedback vectors corresponding to all records in each candidate option subset; perform a statistical significance test on the mean of the experience feedback vectors of different candidate option subsets within the same behavioral feature cluster; The subsequent attraction identifiers corresponding to the subset of candidate options whose mean experience feedback vector is significantly higher than that of other options in the test results are extracted as the experience optimization decision set of the behavioral feature cluster; Establish a mapping relationship between the behavioral feature clusters of the specific scenic spot and the experience optimization decision set; integrate all mapping relationships of all specific scenic spots to construct the association rule base.

[0012] As a further aspect of the present invention: the specific process of initial trip planning in S3 is as follows: Parse the current user request to obtain a list of target attraction identifiers, and generate a set of attractions to be planned; match the set of attractions to be planned with the specific attraction set, and mark the identifiers belonging to the specific attraction set as nodes to be optimized; For each node to be optimized, query its mapped experience optimization decision set from the association rule base, and extract the attraction identifiers contained in the set; add the extracted attraction identifiers to the set of attractions to be planned to form an expanded attraction set; Based on the expanded set of attractions and the spatiotemporal constraint data between attractions, an initial attraction visit sequence is generated; the initial attraction visit sequence is traversed to identify the nodes to be optimized that are arranged as consecutive visit nodes; When consecutive nodes to be optimized are identified, the experience optimization decision set mapped to the first node to be optimized is obtained from the association rule base; Determine whether the second node to be optimized belongs to the scenic spot identifier in the experience optimization decision set; if it does not belong, select a scenic spot identifier from the experience optimization decision set and replace the second node to be optimized in the sequence or insert it before it; The adjusted attraction visit sequence is combined with traffic and time data to generate the initial itinerary plan.

[0013] As a further aspect of the present invention: in step S4, the specific calculation process of the behavioral feature vector is as follows: Obtain the real-time location sequence and dwell time sequence uploaded by the current user's mobile terminal; extract the departure time of the most recently visited attraction from the real-time location sequence, and extract the subsequent real-time location sequence with a fixed duration window starting from that time point; Based on the subsequent real-time positioning sequence within the fixed time window, calculate the movement speed sequence within the fixed time window, and calculate the rate of change of the movement speed sequence compared to the movement speed sequence within the control time window before the departure time point. Based on the dwell time sequence, the frequency of dwell events within the fixed duration window is statistically analyzed, and the rate of change of this frequency relative to the frequency of dwell events within the control duration window is calculated. The calculated rate of change of the movement speed sequence is combined with the rate of change of the dwell event frequency to generate the current user's current behavior feature vector.

[0014] As a further aspect of the present invention: in step S4, the specific construction process of the preset behavior offset classification model is as follows: From the association rule base, extract the cluster center feature vector of each behavioral feature cluster corresponding to all specific attractions; use each cluster center feature vector as a model training sample, and use the identifier of the behavioral feature cluster from which the cluster center feature vector originates as the category label of the training sample; Collect all training samples and category labels corresponding to all the specific attractions to construct a labeled training sample set; use the labeled training sample set to perform supervised training on a multi-class machine learning model with a preset basic structure; The multi-class machine learning model that has been trained and has reached the preset accuracy index is saved as the behavior offset classification model.

[0015] As a further aspect of the present invention: the specific process of dynamically adjusting the remaining travel plans of the current user in step S5 is as follows: Obtain the offset category to which the current user's real-time behavior belongs, as output by the behavior offset classification model; and query the corresponding experience optimization decision set from the association rule base according to the offset category. Obtain the current user's real-time geographic location information and the sequence of attractions to be visited in the remaining itinerary plan; using the real-time geographic location as the starting point, the experience optimization decision set as the candidate destination constraint, and combining real-time traffic data and attraction opening time data, generate subsequent itinerary segments; The generated subsequent itinerary segments are merged and compared with the original attraction sequences that have not been executed in the remaining itinerary plan. If the attraction identifier in the original attraction sequence does not belong to the experience optimization decision set, then the attraction identifier in the subsequent itinerary segments that belongs to the experience optimization decision set is used to replace the attraction identifier that does not belong. The new itinerary sequence, after fusion and replacement processing, is combined with real-time updated traffic movement data and estimated travel time to obtain the adjusted dynamic itinerary plan.

[0016] The beneficial effects of this invention are: 1) By establishing a verifiable correlation between behavioral patterns and experience outcomes, this invention achieves a fundamental shift in itinerary planning from static resource matching to dynamic experience management. Traditional methods rely on fixed attraction tags and user interests, essentially treating tourism as a static resource allocation problem. This invention overcomes this limitation. Its core principle lies in constructing a data analysis framework capable of interpreting group "experience traces." First, it precisely correlates and quantifies tourists' objective behavioral trajectories (such as movement speed and dwell patterns) with their subjective experience feedback (ratings, emotional tendencies), forming coupled "experience-behavior" data. Then, using statistical hypothesis testing methods, it identifies attractions from massive amounts of data that significantly diminish subsequent experiences and precisely characterizes the common behavioral shifts exhibited by tourists after visiting these attractions. Furthermore, within the same behavioral pattern group, through rigorous statistical comparison, it selects subsequent activity choices that bring significantly higher experience evaluations to the group. This ensures that the final decision rules are not based on subjective guesswork or simple frequency, but rather stem from empirical findings on the causal chain of "behavior-choice-outcome." Therefore, the planning scheme generated by this method is based on an optimization strategy validated by historical group experiences, realizing a paradigm shift from arranging "attractions" to managing "experience flows".

[0017] 2) This invention achieves proactive risk avoidance, generating a personalized initial plan with inherent resilience. It injects historically mined experience risk patterns and optimization strategies into the new itinerary generation logic. When planning the initial route for the user, it proactively identifies known high-risk points in the itinerary draft and predicts the user's potential behavioral state after visiting those points. Based on this prediction, it immediately invokes a historically validated set of experience optimization decisions bound to that state from the association rule base, automatically reconstructing the subsequent itinerary—for example, inserting suitable buffer items between high-intensity activities or replacing them with more suitable attractions. Essentially, this process transforms the "lessons learned" from countless historical travelers into a pre-set optimization algorithm, providing avoidance solutions before the user experiences fatigue. Therefore, the final delivered initial plan is a "resilient plan" with a built-in risk-resistant structure, fundamentally improving the scientific rigor and reliability of the plan and significantly reducing the probability of a breakdown in the later stages of the trip due to improper planning.

[0018] 3) This invention constructs an online intelligent closed loop of "real-time perception - state diagnosis - dynamic optimization," providing accompanying personalized services and the ability for sustainable evolution. It extends trip planning from a one-time pre-trip service to a fully intelligent agent covering the entire trip, from pre-trip planning to during-trip activities. This invention collects user behavior data in real time via mobile terminals and uses a pre-trained classification model to quickly match it with a historically accumulated library of typical behavior patterns, thereby achieving an objective diagnosis of the user's real-time fatigue state. Subsequently, based on the diagnosed state category, it retrieves and calls the corresponding optimization decision set, and combines it with real-time traffic and location information to generate an immediate trip adjustment plan and push it to the user. The core value of this closed loop lies not only in achieving dynamic personalization "adapting to the times," but also in constructing a self-evolving pathway. Each real-time diagnosis and subsequent feedback after the user adopts the suggestion will serve as new "experience-behavior" related data, anonymized and fed back to the database for periodic retraining and validation of the rule base and classification model. This allows decision-making knowledge to be continuously updated and evolved, forming a virtuous cycle of "data-driven optimization, and optimization generating new data," thereby continuously enhancing the accuracy and adaptability of its services. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the process of the AI-based travel itinerary planning method of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 As shown, this invention is a tourism itinerary planning method based on artificial intelligence, comprising the following steps: S1. Obtain the historical travel sequences and corresponding travel evaluation data of anonymous user groups in the target area, and construct a travel-evaluation mapping library; First, the application programming interface (API) calls the tourism service platform's backend database to retrieve all anonymous user travel records for the target geographic area (e.g., Huangshan City, Anhui Province) within the past statistical period (e.g., 24 months). Each record is indexed by a unique travel identifier and contains a sequence of attractions visited, precisely sorted by timestamp, for example, "Travel ID-1001: [Timestamp 1: Hongcun, Timestamp 2: Huangshan Scenic Area, Timestamp 3: Xidi]". Simultaneously, the system retrieves evaluation data corresponding to each travel identifier from the evaluation database. This data includes numerical overall ratings (e.g., 4.5 / 5.0) and textual reviews (e.g., "My legs were sore after climbing Huangshan, but the scenery was absolutely stunning").

[0023] Subsequently, automated sentiment analysis was performed on the text comments. A pre-trained natural language processing model was used to identify and extract keywords representing experiential tendencies and their sentiment intensity in the comments. For example, "legs weak" was identified as a negative physiological state label with an intensity value of 0.8 (range 0-1), and "absolutely beautiful" was identified as a positive landscape evaluation label with an intensity value of 0.9. These discrete labels and intensity values ​​were then transformed into a multi-dimensional sentiment vector.

[0024] Finally, the numerical overall rating and the generated sentiment vector are concatenated and normalized to form a comprehensive experience feedback vector. Using the trip identifier as a unique key, the sequence of attractions visited during the trip is bound to its experience feedback vector, generating a complete trip-evaluation mapping record. By collecting a massive number of such records, a trip-evaluation mapping library that can be used for in-depth analysis is constructed.

[0025] S2, perform group behavior analysis on the itinerary-evaluation mapping library, identify common behavior shift feature values ​​after visiting a specific attraction, and cluster the attraction visit records in the historical itinerary sequence that are located after the specific attraction based on the common behavior shift feature values ​​to form an association rule library between behavior shift feature values ​​and subsequent attraction selection clusters. First, specific attractions are identified. By traversing the itinerary-evaluation mapping database, for each participating attraction, the following calculations are performed: For all itinerary records containing the attraction, the itinerary is divided into a "preceding subsequence" and a "subsequence" based on the visit time of the attraction. The mean of the experience feedback vectors corresponding to the preceding and subsequent subsequences is calculated, and their difference is obtained to obtain the "individual experience feedback attenuation degree" for that attraction in a single itinerary. Next, statistical analysis is performed on the attenuation degree of all relevant itineraries (e.g., calculating the mean and variance), and methods such as t-tests are used to determine whether the attenuation is statistically significant (p-value less than 0.05). Attractions that cause statistically significant negative attenuation of experience feedback are marked as specific attractions (i.e., high-risk experience points), such as "Huangshan Scenic Area".

[0026] Secondly, common behavioral shift feature values ​​are calculated. For each specific attraction, several travel records with the most significant "individual experience feedback decay" are selected to form a high-decay record subset. For each record in this subset, GPS trajectory data is extracted within a fixed time window (e.g., the following 24 hours) after the user leaves the specific attraction. By calculating indicators such as the user's average movement speed and number of stops per unit distance within this time window, and comparing them with the baseline values ​​of the previous control time window before visiting the attraction, behavioral shift indicators such as "movement speed change rate" and "stop frequency change rate" are obtained. These indicators are aggregated across all records, and their mean and standard deviation are calculated to collectively constitute common behavioral shift feature values ​​describing the group's behavioral patterns after visiting the specific attraction, such as a vector (speed decay rate: -30%±5%, stop growth rate: +50%±10%).

[0027] Subsequently, an association rule base is constructed. Based on the calculated common behavioral offset feature values, cluster analysis (such as using the K-means algorithm) is performed on the trips within the high decay record subset, and users with similar behavioral patterns are grouped into different behavioral feature clusters, such as "Cluster A: significant deceleration and frequent rest" and "Cluster B: moderate deceleration".

[0028] For each behavioral feature cluster, we analyze what the next attraction visited immediately after a specific attraction is for all itineraries within the cluster. The key step then involves comparing the mean of the experience feedback vectors of subgroups within the cluster that choose different subsequent attractions (e.g., "Hongcun" or "Huizhou Ancient City") to see if there are statistically significant differences. Through hypothesis testing, we identify the subsequent attractions that bring statistically significantly higher experience feedback to that behavioral feature cluster and categorize them into the experience optimization decision set for that cluster.

[0029] Ultimately, a mapping rule is established such as "specific scenic spot (Huangshan) + behavioral feature cluster (cluster A) → experience optimization decision set ({Huizhou Ancient City, hot spring hotel})", and all rules are integrated to form an association rule base.

[0030] S3, responding to the current user's planning request, and generating an initial itinerary plan containing a sequence of attraction visits based on the association rule base; Upon receiving the target attraction list submitted by the current user (e.g., "Huangshan, Hongcun, Xidi"), the system first compares it with a specific attraction set to identify high-risk points as nodes to be optimized (e.g., "Huangshan"). Based on the nodes to be optimized, the system queries the association rule base to pre-obtain candidate attractions from the corresponding experience optimization decision set (e.g., "Huizhou Ancient City") and adds them to the candidate list, forming an expanded attraction set.

[0031] Then, based on the expanded set of attractions, and taking into account the spatiotemporal constraints such as the distance between attractions, opening hours, and typical visit duration, a preliminary sequence of attractions is generated using a path planning algorithm (such as a constrained traveling salesman problem algorithm).

[0032] After generation, a "risk simulation" check is performed on the sequence: It scans the sequence to see if any nodes to be optimized are consecutively arranged (e.g., "Huangshan" immediately followed by "Hongcun"). If so, it calls the experience optimization decision set corresponding to the first node (Huangshan) in the rule base and checks if the second node (Hongcun) is included. If not, it automatically selects a scenic spot (e.g., Huizhou Ancient City) from the experience optimization decision set and inserts it between the two high-risk points as a buffer, or replaces the second node, thus generating an initial itinerary plan with an embedded fatigue risk avoidance strategy.

[0033] S4, during the process of the current user executing the initial itinerary planning, the user's location sequence and dwell time sequence are obtained in real time and the behavior feature vector is calculated. The behavior feature vector is input into the preset behavior offset classification model and the offset category to which the current user's real-time behavior belongs is output. During the trip execution phase, the user-authorized mobile terminal application continuously collects encrypted GPS location sequence and the duration of stay at each location.

[0034] When a user finishes visiting a specific scenic spot (such as Huangshan), location and dwell time data are extracted within a fixed time window (such as 6 hours) starting from the departure time. Based on this data, indicators such as average movement speed and dwell time frequency within this time window are calculated and compared with the baseline values ​​of the user's visit to the scenic spot in the previous time window. The "speed change rate" and "dwell time frequency change rate" are calculated in real time to form a behavioral feature vector describing the current state.

[0035] Previously, a multi-class behavior offset classification model (such as a support vector machine or neural network classifier) ​​had been trained based on the cluster center vectors of all behavior feature clusters in the association rule base. Now, the real-time calculated behavior feature vector is input into the model, and the model will output which historical behavior feature cluster (such as "cluster A: significant deceleration and frequent rest") it best matches, that is, determine the user's current real-time behavior offset category.

[0036] S5. Obtain the corresponding target attraction selection cluster from the association rule base according to the offset category, dynamically adjust the remaining itinerary plan of the current user based on the target attraction selection cluster, and synchronize it to the current user's terminal device.

[0037] Based on the offset category determined by S4, the system immediately queries the association rule base to retrieve the experience optimization decision set mapped to that category (e.g., for "Cluster A", the decision set is {Huizhou Ancient City, Hot Spring Hotel}). Simultaneously, it obtains the user's real-time geolocation and remaining unexecuted itinerary planning. Starting from the user's current location and using attractions in the experience optimization decision set as candidate destinations, it re-plans and generates one or more optimal "subsequent itinerary segments" based on real-time traffic conditions, attraction opening status, and other information.

[0038] The newly generated "following itinerary segment" is compared with the original plan. If the next attraction to be visited in the original plan is not in the experience optimization decision set, it will be replaced or inserted using attractions from the new segment. For example, the original plan was to visit "Hongcun" tomorrow, but based on real-time status judgment and the rule base, it will be suggested to visit "Huizhou Ancient City" instead. Finally, the dynamically adjusted new itinerary, along with updated traffic guidance and time suggestions, is pushed to the user's terminal device interface in real time, completing a personalized itinerary optimization based on real-time physiological status feedback.

[0039] In a preferred embodiment of the present invention, the specific construction process of the trip-evaluation mapping library in step S1 is as follows: Obtain the historical itinerary records and historical evaluation records of anonymous user groups; parse out the attraction visit sequence indexed by itinerary identifiers from the historical itinerary records, the attraction visit sequence containing attraction identifiers arranged by timestamps; The evaluation data corresponding to each itinerary identifier is parsed from the historical evaluation record set. The evaluation data includes numerical overall score data and text-based comment data. Sentiment analysis is performed on the text-based comment data to extract the sentiment tags and intensity values ​​that represent experience tendencies and generate sentiment vectors. The numerical overall rating data is concatenated with the emotion vector to form an experience feedback vector corresponding to the itinerary identifier; the itinerary identifier is used as the association key to bind the attraction visit sequence with the experience feedback vector corresponding to the itinerary identifier, generating an itinerary-evaluation mapping record; all itinerary-evaluation mapping records corresponding to all itinerary identifiers are collected to generate an itinerary-evaluation mapping library.

[0040] First, two fundamental datasets need to be acquired: one is a set of itinerary records documenting countless tourists' past journeys, and the other is a set of evaluation records storing their experiences of these journeys. This operation is necessary because only by placing objective behavior and subjective feedback within the same analytical framework can we discern the patterns behind the experiences. For example, from the itinerary records, we can extract records indexed by unique identifiers such as "Itinerary ID-1001." Each record contains a series of attraction markers precisely arranged chronologically, such as "08:00 Hongcun Ticket Office, 10:30 Huangshan Yunggu Cableway, 14:00 Welcoming Pine…", which reconstructs the tourists' movement trajectories in both spatial and temporal dimensions. Simultaneously, from the evaluation records, we can locate the evaluation data corresponding to the same "Itinerary ID-1001," which includes both overall numerical ratings such as "4.5 stars" and textual reviews such as "Huangshan's scenery is breathtaking, but climbing is incredibly tiring; my legs are practically giving out." The reason why it is necessary to analyze and connect these two is that a single rating cannot reveal the specific reasons for the "tiredness", while isolated comments are detached from the specific context of the trip. Combining the two can lay the foundation for subsequent analysis.

[0041] Next, in-depth interpretation of unstructured text comments is needed, namely sentiment analysis. The principle is to use natural language processing (NLP) technology to enable computers to understand the emotional tone and evaluation objects in human language. For example, analyzing the above comments, we can identify "shocking" as a positive emotion, associated with "scenery"; and "tired" as a negative physiological feeling, associated with "mountain climbing." Furthermore, these identified emotions are assigned intensity values, and this scattered "emotion-object-intensity" information is integrated and encoded into a multi-dimensional sentiment vector. The purpose of this is to transform abstract, difficult-to-calculate subjective feelings into structured, quantifiable, and comparable numerical features, thereby making the "experience" measurable and analyzable.

[0042] Then, the numerical rating representing overall satisfaction is concatenated and fused with the emotional vector depicting detailed feelings to form a comprehensive, multi-dimensional experience feedback vector. This vector integrates overall satisfaction and fine-grained emotional dimensions. The reason for this combination is that the overall rating is the final result of the experience, while the emotional vector reveals the various intrinsic factors that constitute this result; combining the two allows for a more refined representation of experience quality. After this step, using "Trip ID-1001" as a unique key, the previously parsed attraction visit time series is bound to the unique experience feedback vector of this trip, generating a complete record containing "behavioral trajectory" and "experience profile"—a trip-evaluation mapping record. When millions of such records are collected, a trip-evaluation mapping library for in-depth analysis is constructed. By creating a knowledge base capable of large-scale, computable correlation between the "behavioral patterns" and "experience results" of a massive user group, it allows us to move beyond viewing attraction popularity or user ratings in isolation and instead study the core question: "How do tourists actually feel under a specific tour sequence and pace?" For example, the database might contain itinerary patterns like "visiting Hongcun immediately after visiting Huangshan," where the "fatigue" dimension is significantly higher in the corresponding experience feedback vector; while the pattern of "resting at a hot spring hotel after Huangshan" shows more emphasis on the "relaxation" and "overall satisfaction" dimensions. The establishment of this database provides indispensable data and validation for subsequent identification of attractions prone to fatigue and discovery of itinerary optimization strategies to improve satisfaction. It serves as the underlying data pillar for the entire solution to achieve the leap from "planning itineraries" to "managing experiences."

[0043] In another preferred embodiment of the present invention, the specific process of obtaining a specific scenic spot in step S2 is as follows: Extract all attraction visit sequences from the trip-evaluation mapping database, and aggregate the attraction identifiers to form an initial attraction set; for each attraction identifier in the initial attraction set, retrieve all trip-evaluation mapping records containing that identifier from the trip-evaluation mapping database to form an associated record set; For each record in the associated record set, locate the visit time point of the attraction identifier in its attraction visit sequence, divide the sequence into a preceding subsequence and a subsequent subsequence based on the time point; obtain the mean local experience feedback corresponding to the preceding subsequence and the subsequent subsequence respectively, and calculate the difference between the two as the individual experience feedback attenuation degree of the record. Statistical analysis is performed on the individual experience feedback decay rate of all records in the associated record set to obtain the group experience feedback decay rate and its statistical distribution parameters of the scenic spot identifier; after traversing the initial scenic spot set, the group experience feedback decay rate and its statistical distribution parameters of each scenic spot identifier are filtered through statistical significance test. Scenic spots with negative group experience feedback decay and statistical significance exceeding a preset threshold are marked as specific scenic spots, forming a specific scenic spot set.

[0044] The system extracts all recorded travel routes and compiles all attraction names appearing on each route, forming a large "initial attraction set." This set might include hundreds or even thousands of locations such as "Huangshan," "Hongcun," "Xidi," and "Huizhou Ancient City." This is done because it requires an independent and fair evaluation of each attraction within a unified framework, without pre-assuming any attraction attributes. Next, for each attraction in the set, such as "Huangshan," a search is performed across the entire mapping database to find all records where "Huangshan" appears in tourist journeys. These records constitute the "association record set" for "Huangshan." The principle behind this step is that to scientifically assess the impact of an attraction, it must be based on a sufficient number of real-world cases involving that attraction, avoiding misjudgments caused by isolated, accidental events.

[0045] Subsequently, each personal journey record in the "Associated Record Set" was analyzed in depth. Taking a tourist record containing "Huangshan" as an example, the specific time when "Huangshan" was visited was first precisely located on the timeline of this journey. Then, using this time point as a boundary, the entire journey was divided into "the part before Huangshan" and "the part after Huangshan," i.e., the preceding subsequence and the subsequent subsequence. Next, the "experience feedback vector" associated with these two subsequences was read. Since the "experience feedback vector" is a comprehensive numerical representation that includes overall rating and multiple emotional dimensions, by calculating the average of all its dimensions, a value representing the "average experience level before Huangshan" and a value representing the "average experience level after Huangshan" can be obtained. Subtracting the "before" value from the "after" value yields a difference, which is defined as the "individual experience feedback decay degree" of "Huangshan" in this record. If the difference is negative, it means that the tourist's experience level decreased after visiting Huangshan. The principle behind this is that by comparing the changes in the state of the same individual before and after a specific event, the influence of the event itself can be more effectively isolated, reducing the interference caused by differences in the basic experiences of different individual tourists.

[0046] After calculating the individual decay rates of all associated records for "Huangshan," all these positive and negative individual decay rate values ​​are grouped together, and their average (i.e., the group experience feedback decay rate) and fluctuation range (statistical distribution parameter) are calculated. The average reflects the overall trend of "Huangshan's" impact on the tourist experience, while the fluctuation range indicates whether this impact is stable and consistent. After traversing all attractions in the "initial attraction set," the list is finally filtered using statistical significance tests. For example, for "Huangshan," a statistical test is conducted to determine whether the calculated negative average decay rate is large enough to be unlikely to be caused by random fluctuations; similarly, "Huizhou Ancient City" may also show a slight negative decay, but its fluctuations may be large enough that the statistical test cannot confirm its significance. Only those attractions, like "Huangshan," whose group experience feedback decay rate is not only negative (meaning it generally leads to a decline in experience), but whose negative impact is confirmed as "significant" through statistical testing, are ultimately marked as "specific attractions" and form the "specific attraction set." The fundamental principle behind this approach is to introduce scientific statistical inference, transforming the subjective feeling of "being tired at a certain scenic spot" into objective and verifiable data conclusions, thereby ensuring that the identified risk points are real and widespread, rather than accidental or speculative.

[0047] Traditional methods may rely on operational experience or sporadic complaints to guess which attractions are overly demanding, but this is neither comprehensive nor accurate. This invention, through the steps described above, can proactively answer the crucial question: "Which attractions genuinely and universally lower tourists' subsequent experience?" The advantage of this is that it provides precise, data-driven "targets" for all subsequent optimization strategies. Identifying "Huangshan" as a "specific attraction" and placing it in a "set of specific attractions" means gaining crucial knowledge: the arrangement of "Huangshan" in itinerary planning requires exceptional care and optimized design. Only by knowing where the "problem" is most likely to occur can preventative planning be effectively implemented during the planning stage, such as avoiding a tight arrangement of multiple "specific attractions" or inserting buffer zones between them. Therefore, this step is an indispensable bridge connecting "historical data phenomenon discovery" and "proactive optimization of future itineraries," elevating the entire method from passively recording feedback to proactively predicting and managing risks.

[0048] In another preferred embodiment of the present invention, the specific process of the common behavior offset feature value in S2 is as follows: For each specific attraction in the specific attraction set, extract its corresponding associated record set from the itinerary-evaluation mapping library; filter out records from the associated record set whose individual experience feedback decay is negative and whose absolute value exceeds a first preset threshold to form a high decay record subset; For each record in the high-attenuation record subset, the trajectory data within a fixed time window starting from the visit time of the specific scenic spot in its scenic spot visit sequence is extracted; based on the trajectory data, the rate of change of movement speed and the rate of change of dwell event frequency of the record within the fixed time window compared to a control time window before the visit time point are calculated; The rate of change of movement speed and the rate of change of dwell event frequency of all records in the high-attenuation record subset are collected, and their arithmetic mean and standard deviation are calculated respectively. The combination of the arithmetic mean and standard deviation of the rate of change of movement speed and the arithmetic mean and standard deviation of the rate of change of dwell event frequency is used as the common behavior offset feature value corresponding to the specific scenic spot.

[0049] First, for each identified "specific attraction," such as "Huangshan," a secondary screening is performed from its vast "associated record set." Instead of analyzing all tourists who have visited Huangshan, the focus is on selecting records with negative "individual experience feedback attenuation" and large absolute values—those whose experiences after visiting Huangshan clearly and significantly worsened. This constitutes a "high-attenuation record subset." By analyzing the group with the most significant experience impairment, we can more clearly capture the potentially common external behavioral manifestations accompanying the decline in experience, eliminating interference from individual cases where the experience of visiting Huangshan was not significantly affected or even improved. This makes the subsequent search for behavioral patterns more representative and targeted.

[0050] Next, micro-behavioral analysis is performed on each record in the subset. Taking a selected tourist record as an example, a fixed, statistically significant time window (e.g., the next 24 hours) is extracted from its attraction visit sequence, starting from the precise time of departure from "Huangshan". Simultaneously, to establish a comparable baseline, trajectory data from a "control time window" of the same length prior to the tourist's visit to "Huangshan" is also extracted. Based on continuous location coordinates (e.g., GPS track points) and timestamps within these two time windows, calculations are performed: the average movement speed is calculated by analyzing the distance of coordinate change per unit time, and a "stay event" is defined by identifying track points clustered in one place for an extended period. Then, the average movement speed within the fixed time window is divided by the average movement speed within the control time window, and then 1 is subtracted to obtain the "movement speed change rate". Similarly, the number of stay events per hour within the fixed time window is subtracted from the number per hour within the control time window, and then divided by the control value to obtain the "stay event frequency change rate". It is understandable that a decrease in walking speed can directly reflect a decline in physical strength and a reduction in the willingness to move, while an abnormally high frequency of stops may reflect a physiological need for more rest or mental fatigue and hesitation. By calculating the rate of change rather than the absolute value, the differences in walking speed among different tourists can be eliminated, focusing on measuring the relative behavioral changes of the same individual before and after a specific event (visiting Huangshan).

[0051] After completing the above calculations for tourists within all high-attenuation subsets, the inductive abstraction stage begins. This involves aggregating the "rate of change in movement speed" calculated for all tourists and calculating their arithmetic mean and standard deviation. The mean represents the average percentage decrease in movement speed for tourists whose experience significantly declined after visiting Huangshan, revealing the general intensity of fatigue; while the standard deviation reflects the range of fluctuation in this speed reduction, indicating the dispersion of fatigue manifestations. The same calculation is performed for the "rate of change in frequency of stay events." Finally, the combination of the mean and standard deviation of these two indicators is used as the "common behavioral deviation characteristic value" corresponding to the specific scenic spot "Huangshan." For example, a possible value is "rate of change in speed: mean -30%, standard deviation 5%; rate of change in frequency of stay events: mean +50%, standard deviation 15%." The principle behind this characteristic value is that it is no longer an isolated number, but a statistical description containing both central tendency and dispersion, which can more robustly and comprehensively define a class of behavioral patterns. It means that a typical tourist whose experience is impaired after visiting Huangshan has a high probability that their subsequent behavior will fall within the range defined by this statistical description.

[0052] By establishing a strong correlation between subjective "deterioration in experience" and objective, measurable "behavioral signals," a clear "fingerprint" or "template" is provided for machine recognition. It becomes clear that "Huangshan" is not only a statistically significant location leading to a decline in ratings, but more specifically, this decline is typically accompanied by quantifiable behavioral patterns such as "a general decrease in speed of about 30% and an increase in rest frequency of about 50%." This approach offers two advantages: First, it enables an indirect, objective, and scalable measurement of "fatigue" or "impaired experience." Without asking users "Are you tired?", probabilistic inferences can be made by analyzing their anonymous trajectory data. When a new tourist visits Huangshan, their actual behavioral indicators can be calculated and compared with this pre-prepared "common behavioral deviation feature value" to determine whether the tourist is currently entering a similar state of declining experience experienced by many in the past. This completes the key knowledge transformation from identifying problems (specific scenic spots) to representing problems (behavioral deviation features), and then to enabling real-time diagnosis of the problems. Without this step, it's impossible to evolve from "knowing that Huangshan might be tiring" to "being able to identify tourists who are getting tired," and all subsequent personalized dynamic interventions would be out of the question, with the solution remaining at the level of post-event analysis.

[0053] In another preferred embodiment of the present invention, the specific construction process of the association rule base in step S2 is as follows: For each specific attraction in the set of specific attractions, obtain its high-attenuation record subset; perform cluster analysis based on the common behavioral offset feature values ​​of each record in the high-attenuation record subset to form multiple behavioral feature clusters; For each record in the high-decay record subset, extract the identifier of the first subsequent attraction following that specific attraction in its attraction visit sequence and its corresponding experience feedback vector; The subsequent attraction identifiers and the experience feedback vectors are assigned to the corresponding behavioral feature clusters; for each behavioral feature cluster, the records within it are grouped according to the subsequent attraction identifiers to form multiple candidate option subsets; Calculate the mean of the experience feedback vectors corresponding to all records in each candidate option subset; perform a statistical significance test on the mean of the experience feedback vectors of different candidate option subsets within the same behavioral feature cluster; The subsequent attraction identifiers corresponding to the subset of candidate options whose mean experience feedback vector is significantly higher than that of other options in the test results are extracted as the experience optimization decision set of the behavioral feature cluster; Establish a mapping relationship between the behavioral feature clusters of the specific scenic spot and the experience optimization decision set; integrate all mapping relationships of all specific scenic spots to construct the association rule base.

[0054] First, for each labeled "specific attraction," such as "Huangshan," its corresponding "high-attenuation record subset" is retrieved—that is, the records of tourists whose experience significantly declined after visiting Huangshan. The "common behavioral shift feature values" (such as a set of values ​​containing changes in speed and dwell time) carried by each of these records are analyzed. Based on the similarity of these values, cluster analysis is used to automatically divide all records into several intrinsically consistent groups, namely "behavioral feature clusters." For example, it might be found that among these tourists with poor experiences, some exhibit behavioral characteristics of "a sharp decrease in movement speed and very frequent rest," and are classified as "high fatigue cluster A"; while others show "a moderate decrease in speed and a slight increase in rest," and are classified as "medium fatigue cluster B." It is understandable that even with the same decline in experience, the underlying fatigue levels and performance patterns may differ. Refined clustering helps to provide more accurate and tiered optimization suggestions, rather than giving a general solution for all situations.

[0055] After clustering, the study explores the choices made by tourists in each cluster under fatigue and their consequences. It iterates through each record in the "high-decay record subset," locating the next place a tourist visits after leaving a "specific attraction" (Huangshan), and extracting the identifier of that place, such as "Hongcun" or "Huizhou Ancient City." Simultaneously, it reads the complete "experience feedback vector" corresponding to this record, representing the tourist's final evaluation and feelings about the entire trip. Subsequently, this pair of information—"subsequent attraction identifier" and "experience feedback vector"—is categorized under the "behavioral feature cluster" to which the record belongs. Establishing the correlation between "behavioral state," "subsequent actions," and "final experience" aims to answer a key question: When in a specific fatigue state (e.g., high-fatigue cluster A), do different subsequent activities (going to Hongcun or Huizhou Ancient City) result in different experience outcomes? Detailed statistical analysis was performed within each "behavioral feature cluster." First, all records within the cluster were grouped according to their chosen "subsequent attraction identifiers." For example, in the "high fatigue cluster A," multiple "candidate option subsets" were formed, such as the group that chose Hongcun and the group that chose Huizhou Ancient City. Next, the average of all "experience feedback vectors" within each subset was calculated. This average quantitatively reflects the overall experience level of the group that chose that option. Then, statistical significance tests (e.g., t-tests) were performed between these subsets to mathematically determine whether the differences in the mean experience from different options were large enough to be unlikely to be due to random luck. For example, the test might find that in the "high fatigue cluster A," the mean experience feedback of the "group that chose Huizhou Ancient City" was significantly, rather than by chance, higher than the mean of the "group that chose Hongcun." Finally, the attraction identifiers (such as "Huizhou Ancient City") corresponding to the option (or those that) brought statistically significantly higher experience feedback were formally extracted as the "experience optimization decision set" for that behavioral feature cluster. This means that historical data has validated a rule of thumb: "When tourists are in a state of high fatigue (A), going to Huizhou Ancient City instead of going to Hongcun as originally planned will statistically result in a better overall experience." Finally, a mapping relationship is established between each "specific attraction" and its various "behavioral characteristic clusters" and corresponding "experience optimization decision sets" (e.g., "Huangshan -> High Fatigue Cluster A -> Recommended {Huizhou Ancient City}"). Integrating these mapping relationships for all specific attractions constitutes the "association rule base" driving the entire intelligent planning process. This distills massive amounts of fuzzy historical experience into clear, verifiable, and actionable decision knowledge. Its purpose is not simply to record where tourists went, but to discover and solidify best practices for "what actions should be taken in what state to achieve better results." This rule base is the "brain" and "knowledge base" of intelligent decision-making. It enables the system to move beyond simple geographical and popularity logic when planning initial itineraries or making dynamic adjustments, instead relying on empirically proven experience optimization strategies that match physiological states to make recommendations. For example, when it anticipates or detects that a tourist is about to or has already entered a "high fatigue state A," it will unhesitatingly retrieve the optimization decision "Huizhou Ancient City" from the rule base, rather than recommending a popular attraction that might exacerbate fatigue.

[0056] In another preferred embodiment of the present invention, the specific process of initial trip planning in step S3 is as follows: Parse the current user request to obtain a list of target attraction identifiers, and generate a set of attractions to be planned; match the set of attractions to be planned with the specific attraction set, and mark the identifiers belonging to the specific attraction set as nodes to be optimized; For each node to be optimized, query its mapped experience optimization decision set from the association rule base, and extract the attraction identifiers contained in the set; add the extracted attraction identifiers to the set of attractions to be planned to form an expanded attraction set; Based on the expanded set of attractions and the spatiotemporal constraint data between attractions, an initial attraction visit sequence is generated; the initial attraction visit sequence is traversed to identify the nodes to be optimized that are arranged as consecutive visit nodes; When consecutive nodes to be optimized are identified, the experience optimization decision set mapped to the first node to be optimized is obtained from the association rule base; Determine whether the second node to be optimized belongs to the scenic spot identifier in the experience optimization decision set; if it does not belong, select a scenic spot identifier from the experience optimization decision set and replace the second node to be optimized in the sequence or insert it before it; The adjusted attraction visit sequence is combined with traffic and time data to generate the initial itinerary plan.

[0057] When a user enters a list of desired tourist attractions, such as "Huangshan, Hongcun, and Xidi," into the application interface, the request is first parsed and transformed into a machine-processable "set of attractions to be planned." By standardizing the user's free-form language into structured data objects, a clear input range is established for all subsequent automated decisions. Next, this set of user preferences is compared with a "set of specific attractions" derived from historical data analysis. For example, if "Huangshan" is found in the specific attraction set, it is immediately marked as a "node to be optimized" in the set of attractions to be planned. This is understandable, as historical knowledge indicates that "Huangshan" is an attraction requiring careful planning and prone to fatigue accumulation; therefore, it is highlighted at the outset for focused attention.

[0058] After identifying all nodes to be optimized, for each node, such as "Huangshan," it queries historical data to determine which subsequent choices have proven to lead to a better experience when tourists are in different fatigue states caused by this attraction. It extracts the corresponding "experience optimization decision set" from the rule base, such as a suggestion set like "{Huizhou Ancient City}". Subsequently, these high-quality alternative attraction identifiers extracted from historical experience (such as "Huizhou Ancient City") are added to the user's original set of preferences, thus forming an "expanded attraction set." The principle is that initial planning should not be limited to the finite options proposed by the user, but should proactively introduce historically validated buffers or alternative resources to expand the solution space for generating better solutions.

[0059] Next, based on the "expanded set of attractions," and combined with "spatiotemporal constraint data" such as the actual distance between attractions, typical visit duration, and opening hours, a path planning algorithm is run to generate a preliminary, spatiotemporally feasible "initial attraction visit sequence." A possible generated result is "Day 1: Hongcun, Day 2: Huangshan, Day 3: Xidi." After generation, a crucial "risk flow simulation" check is performed on this sequence: it traverses the entire sequence, specifically examining whether the marked "nodes to be optimized" are arranged too densely, i.e., checking if two or more such nodes are arranged for consecutive visits. For example, it will find that "Huangshan" and "Xidi" are adjacent in this sequence. Understandably, consecutively visiting multiple high-fatigue-risk points is a typical trap leading to a breakdown in the experience, and this dangerous pattern must be actively investigated.

[0060] Once such consecutive risk nodes are identified, automatic optimization logic is triggered. It first focuses on the first node to be optimized (Huangshan) and retrieves its corresponding "experience optimization decision set" from the association rule base. Then, it determines whether the second node to be optimized (Xidi) is within this optimization decision set. In this example, if "Xidi" is not in the set "{Huizhou Ancient City}", it means that, based on historical experience, visiting "Xidi" immediately after the fatigue of "Huangshan" is not the optimal choice. Therefore, a scenic spot (such as Huizhou Ancient City) is selected from the optimization decision set, and a surgical adjustment is made to the original sequence: either "Huizhou Ancient City" directly replaces "Xidi", or "Huizhou Ancient City" is inserted between "Huangshan" and "Xidi". By using data-driven rules, proactive intervention is made in identified high-risk structures, using historically validated high-quality choices to buffer or replace arrangements that may cause a decline in experience.

[0061] Finally, the adjusted new attraction visit sequence is populated with detailed data such as transportation connections, transfer times, and suggested visit durations, generating an "initial itinerary plan" with a timetable and action guidelines that can be directly delivered to users. This transforms reactive, post-event experience management into proactive, pre-event risk mitigation. Traditional planning methods only connect resources based on users' explicit preferences, blindly ignoring potential experience risks. The process described in this invention simulates the "fatigue load" of the itinerary during the planning stage and predicts and automatically reorganizes risk points based on historical collective wisdom (association rule base). Its aim is to inject "resilience" into the itinerary before it begins, ensuring the generated plan itself has built-in optimization strategies to combat common experience pitfalls.

[0062] In another preferred embodiment of the present invention, the specific calculation process of the behavior feature vector in step S4 is as follows: Obtain the real-time location sequence and dwell time sequence uploaded by the current user's mobile terminal; extract the departure time of the most recently visited attraction from the real-time location sequence, and extract the subsequent real-time location sequence with a fixed duration window starting from that time point; Based on the subsequent real-time positioning sequence within the fixed time window, calculate the movement speed sequence within the fixed time window, and calculate the rate of change of the movement speed sequence compared to the movement speed sequence within the control time window before the departure time point; Based on the dwell time sequence, the frequency of dwell events within the fixed duration window is statistically analyzed, and the rate of change of this frequency relative to the frequency of dwell events within the control duration window is calculated. The calculated rate of change of the movement speed sequence is combined with the rate of change of the dwell event frequency to generate the current user's current behavior feature vector.

[0063] As users move during their journey, applications on their smartphones or other mobile devices continuously and anonymously record two core data streams in the background: first, latitude and longitude coordinates collected at fixed time intervals, forming a "real-time location sequence"; and second, the length of time the user remains stationary at each coordinate point, forming a "stay time sequence." The principle behind collecting this data is that a person's movement patterns in large-scale spaces (such as between scenic spots) and their micro-level pausing habits are the most direct and objective external reflections of their physical condition, level of interest, and willingness to act. Moreover, this information can be obtained imperceptibly through the device's built-in sensors, without requiring the user to actively report it.

[0064] Next, a key reference time point needs to be determined, such as the departure time of the most recently visited attraction. The real-time location sequence will be compared with the known geofence boundaries of the attractions. For example, when it is detected that a user's consecutive coordinate points move from inside the preset geofence of "Huangshan South Gate" to outside the geofence and do not return, it can be determined that the user has left the "Huangshan" attraction, and the timestamp of the last coordinate point within the geofence is recorded as the "departure time point." The principle behind this is that using the moment a user leaves a specific activity node (attraction) as the baseline starting point for evaluating its subsequent state changes ensures that the time anchor points of the analysis have clear physical meaning and consistency.

[0065] Once the departure time is determined, all location data within a predefined "fixed-length window" (e.g., the next 6 hours) is immediately extracted forward (in the future direction) from that time point as the "subsequent real-time location sequence." Simultaneously, to establish a comparable individualized baseline, location data within a "comparison window" (e.g., the last 6 hours of the Huangshan tour) of identical length is extracted forward (in the past direction) from the same departure time point as the endpoint. The principle behind selecting equally long and adjacent windows for comparison is that it maximizes the fairness of the comparison: it compares the behavioral differences of the same user at the end of their visit to the same core attraction (Huangshan) versus the initial stage immediately after their visit. This allows for a purer analysis of the immediate impact of the "completion of the tour" event on their behavioral patterns, reducing confounding factors from different times of day (e.g., mornings are generally more productive than afternoons) or different activity types.

[0066] Subsequently, quantitative calculations are performed based on the positioning sequences from these two time windows. For the sequence within the "control time window," the average movement speed of the user during the later stages of visiting the attraction (such as the final stage of climbing or descending a mountain) is calculated, which can be called the "baseline speed." For the sequence within the "fixed time window" (i.e., the subsequent observation window), the system calculates the average movement speed of the user after leaving the attraction (such as on the way back to the hotel or during evening activities), which is called the "observation speed." Dividing the "observation speed" by the "baseline speed" and then subtracting 1 yields a ratio value, which is the "rate of change of movement speed." If this value is negative, it indicates that the user's movement has slowed down. The principle is that muscle fatigue, mental relaxation, or decreased interest can directly lead to a decrease in a person's displacement efficiency, and the rate of change of speed is a continuous indicator that is sensitive to the overall physiological and psychological energy level.

[0067] Simultaneously, the "stay duration sequence" is analyzed. Within the two identical time windows mentioned above, a criterion for determining a "stay event" is defined (e.g., remaining stationary for more than 90 seconds within a 50-meter diameter). The total number of stay events occurring within the "control window" is counted and divided by the window duration to obtain the "baseline stay frequency." Similarly, the total number of stay events occurring within the "subsequent observation window" is counted and divided by the window duration to obtain the "observed stay frequency." Likewise, the rate of change of the "observed stay frequency" relative to the "baseline stay frequency" is calculated. The principle is that during non-goal-oriented movement (such as the return trip after a tour), an increase in the number of unpurposed and unnecessary stays is often a signal that the body is seeking rest, attention is scattered, or decision-making is slow (e.g., not knowing where to go next), and is closely related to fatigue accumulation.

[0068] Finally, the calculated scalar values ​​of "rate of change in movement speed" and "rate of change in dwell frequency" are combined into a two-dimensional "behavioral feature vector." For example, a possible vector is [-0.25, +0.40], representing a 25% decrease in speed and a 40% increase in dwell frequency. The principle behind this is that a single indicator may be affected by specific contexts (e.g., slow speed might be due to taking photos), but combining two indicators reflecting fatigue from different perspectives (action efficiency and rest needs) can corroborate each other, forming a more robust and comprehensive digital characterization of the user's current overall behavioral state. This vector is no longer a string of raw, ambiguous coordinate points, but a standardized mathematical representation that encapsulates key behavioral change information.

[0069] Understandably, traditional methods cannot determine whether a user is truly "tired" after visiting a tourist attraction. This method, however, transforms the abstract concept of "tiredness" into a concrete, comparable numerical vector through a sophisticated comparison window design and dual-indicator calculation. This enables an indirect yet measurable real-time diagnosis of user fatigue. It does not rely on subjective user statements (which may be inaccurate or undisclosed) or specialized medical devices (such as heart rate monitors), utilizing only readily available smartphone data. This "behavioral feature vector" is a crucial "sensor signal." It serves as a bridge connecting "user's real-world behavior" with the internal "intelligent decision-making model." Subsequently, a classification model identifies the fatigue category to which this vector belongs, triggering corresponding itinerary adjustments.

[0070] In another preferred embodiment of the present invention, the specific construction process of the preset behavior offset classification model in step S4 is as follows: From the association rule base, extract the cluster center feature vector of each behavioral feature cluster corresponding to all specific attractions; use each cluster center feature vector as a model training sample, and use the identifier of the behavioral feature cluster from which the cluster center feature vector originates as the category label of the training sample; Collect all training samples and category labels corresponding to all the specific attractions to construct a labeled training sample set; use the labeled training sample set to perform supervised training on a multi-class machine learning model with a preset basic structure; The multi-class machine learning model that has been trained and has reached the preset accuracy index is saved as the behavior offset classification model.

[0071] In another preferred embodiment of the present invention, the specific process of dynamically adjusting the remaining travel plan of the current user in step S5 is as follows: Obtain the offset category to which the current user's real-time behavior belongs, as output by the behavior offset classification model; and query the corresponding experience optimization decision set from the association rule base according to the offset category. Obtain the current user's real-time geographic location information and the sequence of attractions to be visited in the remaining itinerary plan; using the real-time geographic location as the starting point, the experience optimization decision set as the candidate destination constraint, and combining real-time traffic data and attraction opening time data, generate subsequent itinerary segments; The generated subsequent itinerary segments are merged and compared with the original attraction sequences that have not been executed in the remaining itinerary plan. If the attraction identifier in the original attraction sequence does not belong to the experience optimization decision set, then the attraction identifier in the subsequent itinerary segments that belongs to the experience optimization decision set is used to replace the attraction identifier that does not belong. The new itinerary sequence, after fusion and replacement processing, is combined with real-time updated traffic movement data and estimated travel time to obtain the adjusted dynamic itinerary plan.

[0072] While a user is executing their initial itinerary, such as having just finished visiting Huangshan and checking into a hotel, their mobile application continuously collects and uploads anonymized location and dwell time data in the background. A pre-trained "behavioral offset classification model" analyzes this fresh data in real time. The model works by having learned and stored historical fatigue behavior patterns (such as "high fatigue cluster A" and "medium fatigue cluster B") exhibited by tourists after visiting various "specific attractions" as classification criteria. The model then compares the user's newly generated behavioral data features (such as abnormally slow movement and frequent short stops) with these criteria, ultimately outputting a judgment, such as "Current behavior belongs to: High fatigue cluster A." Essentially, this step completes a transformation from raw sensor data to higher-order physiological state cognition, allowing the model to "understand" the user's current fatigue level.

[0073] Once a clear "offset category" diagnosis is obtained, it is immediately used as a precise query command and retrieved from the "association rule base." The rule base stores historically validated optimal mappings of "state-action-outcome." Based on the key "high fatigue cluster A," the corresponding "experience optimization decision set" is quickly found, such as a set containing tags like "Ancient Huizhou City" and "Hot Spring Experience." This means that historical big data indicates that tourists in a similar fatigue state will have a better subsequent experience by choosing these activities. Next, two key pieces of information are simultaneously acquired: first, the user's "real-time geographical location" obtained through their mobile phone's GPS (e.g., a hotel at the foot of Huangshan Mountain); and second, the "remaining sequence of attractions to be visited" in the user's itinerary (e.g., the original plan to visit "Hongcun" tomorrow). The principle behind acquiring these two pieces of information is to initiate a replanning process under the constraints of the real physical world: knowing "where to start" and "where to originally plan to go."

[0074] Starting with the user's "real-time geolocation," the planning process uses all attraction identifiers in the "experience optimization decision set" as strongly constrained candidate destinations—meaning the planning objective must be selected from these validated and excellent options. Simultaneously, real-time traffic data (such as current road congestion from the hotel to each candidate location and available public transportation) and real-time opening hours for each attraction are integrated. Based on all these dynamic inputs, the engine quickly calculates and generates one or more optimal "subsequent itinerary segments." For example, it might generate a segment: "Tomorrow morning at 9:30, take a direct bus from the current hotel to Huizhou Ancient City, with an estimated 3-hour visit." The principle behind this step is to find the most feasible and efficient execution plan at the current time and location, while ensuring the experience optimization objective (visiting attractions within the decision set).

[0075] Next, the intelligent decision fusion stage begins. It compares the newly generated "follow-up itinerary fragment" with the user's pre-defined, yet-to-be-executed "original attraction sequence." Its core logic is to check for conflicts between the original plan and the optimization suggestions. In this example, it's discovered that the next attraction in the original sequence, "Hongcun," is not within the "experience optimization decision set" {Huizhou Ancient City, Hot Spring Experience}. Based on predetermined optimization rules, the replacement logic is immediately initiated: it uses the attraction identifier from the newly generated itinerary fragment that belongs to the decision set (i.e., "Huizhou Ancient City") to replace the attraction identifier from the original sequence that does not belong to the decision set (i.e., "Hongcun"). The principle is that when the historically validated optimal choice is inconsistent with the original plan, the choice that data proves to bring a better experience is prioritized, thus proactively enhancing the user's journey.

[0076] Finally, the new sequence of attractions obtained after the replacement is combined with real-time updated details such as specific transportation methods, departure times, and estimated travel times, and packaged into a complete and immediate "adjusted dynamic itinerary plan," which is then immediately pushed to the user's mobile terminal, achieving a leap from "static planning" to "accompanying intelligent agent." Traditional itinerary planning essentially ends after the plan is delivered, while the dynamic adjustment mechanism described in this invention enables continuous perception of the user's state during the journey and proactive intervention at critical moments (such as when significant fatigue is detected), providing empirically proven and personalized optimization solutions. The advantage of this is that it greatly enhances the adaptability and resilience of the itinerary, allowing it to flexibly respond to unplanned changes in the user's state.

[0077] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A tourism itinerary planning method based on artificial intelligence, characterized in that, Includes the following steps: S1. Obtain the historical travel sequences and corresponding travel evaluation data of anonymous user groups in the target area, and construct a travel-evaluation mapping library; S2, perform group behavior analysis on the itinerary-evaluation mapping library, identify common behavior shift feature values ​​after visiting a specific attraction, and cluster the attraction visit records in the historical itinerary sequence that are located after the specific attraction based on the common behavior shift feature values ​​to form an association rule library between behavior shift feature values ​​and subsequent attraction selection clusters. S3, responding to the current user's planning request, and generating an initial itinerary plan containing a sequence of attraction visits based on the association rule base; S4, during the process of the current user executing the initial itinerary planning, the user's location sequence and dwell time sequence are obtained in real time and the behavior feature vector is calculated. The behavior feature vector is input into the preset behavior offset classification model and the offset category to which the current user's real-time behavior belongs is output. S5. Obtain the corresponding target attraction selection cluster from the association rule base according to the offset category, dynamically adjust the remaining itinerary plan of the current user based on the target attraction selection cluster, and synchronize it to the current user's terminal device.

2. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S1, the specific construction process of the trip-evaluation mapping library is as follows: Obtain the historical itinerary records and historical evaluation records of anonymous user groups; parse out the attraction visit sequence indexed by itinerary identifiers from the historical itinerary records, the attraction visit sequence containing attraction identifiers arranged by timestamps; The evaluation data corresponding to each itinerary identifier is parsed from the historical evaluation record set. The evaluation data includes numerical overall score data and text-based comment data. Sentiment analysis is performed on the text-based comment data to extract the sentiment tags and intensity values ​​that represent experience tendencies and generate sentiment vectors. The numerical overall rating data is concatenated with the emotion vector to form an experience feedback vector corresponding to the itinerary identifier; the itinerary identifier is used as the association key to bind the attraction visit sequence with the experience feedback vector corresponding to the itinerary identifier, generating an itinerary-evaluation mapping record; all itinerary-evaluation mapping records corresponding to all itinerary identifiers are collected to generate an itinerary-evaluation mapping library.

3. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S2, the specific process for obtaining the information about a particular scenic spot is as follows: Extract all attraction visit sequences from the trip-evaluation mapping database, and aggregate the attraction identifiers to form an initial attraction set; for each attraction identifier in the initial attraction set, retrieve all trip-evaluation mapping records containing that identifier from the trip-evaluation mapping database to form an associated record set; For each record in the associated record set, locate the visit time point of the attraction identifier in its attraction visit sequence, divide the sequence into a preceding subsequence and a subsequent subsequence based on the time point; obtain the mean local experience feedback corresponding to the preceding subsequence and the subsequent subsequence respectively, and calculate the difference between the two as the individual experience feedback attenuation degree of the record. Statistical analysis is performed on the individual experience feedback decay rate of all records in the associated record set to obtain the group experience feedback decay rate and its statistical distribution parameters of the scenic spot identifier; after traversing the initial scenic spot set, the group experience feedback decay rate and its statistical distribution parameters of each scenic spot identifier are filtered through statistical significance test. Scenic spots with negative group experience feedback decay and statistical significance exceeding a preset threshold are marked as specific scenic spots, forming a specific scenic spot set.

4. The artificial intelligence-based tourism itinerary planning method according to claim 2, characterized in that, In S2, the specific process of the common behavior offset feature value is as follows: For each specific attraction in the specific attraction set, extract its corresponding associated record set from the itinerary-evaluation mapping library; select records from the associated record set whose individual experience feedback decay is negative and whose absolute value exceeds a first preset threshold to form a high decay record subset; For each record in the high-attenuation record subset, the trajectory data within a fixed time window starting from the visit time of the specific scenic spot in its scenic spot visit sequence is extracted; based on the trajectory data, the rate of change of movement speed and the rate of change of dwell event frequency of the record within the fixed time window compared to a control time window before the visit time point are calculated; The rate of change of movement speed and the rate of change of dwell event frequency of all records in the high-attenuation record subset are collected, and their arithmetic mean and standard deviation are calculated respectively. The combination of the arithmetic mean and standard deviation of the rate of change of movement speed and the arithmetic mean and standard deviation of the rate of change of dwell event frequency is used as the common behavior offset feature value corresponding to the specific scenic spot.

5. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S2, the specific construction process of the association rule base is as follows: For each specific attraction in the set of specific attractions, obtain its high-attenuation record subset; perform cluster analysis based on the common behavioral offset feature values ​​of each record in the high-attenuation record subset to form multiple behavioral feature clusters; For each record in the high-decay record subset, extract the identifier of the first subsequent attraction following that specific attraction in its attraction visit sequence and its corresponding experience feedback vector; The subsequent attraction identifiers and the experience feedback vectors are assigned to the corresponding behavioral feature clusters; for each behavioral feature cluster, the records within it are grouped according to the subsequent attraction identifiers to form multiple candidate option subsets; Calculate the mean of the experience feedback vectors corresponding to all records in each subset of candidate options; Perform a statistical significance test on the mean of the experience feedback vectors of different candidate option subsets within the same behavioral feature cluster; The subsequent attraction identifiers corresponding to the subset of candidate options whose mean experience feedback vector is significantly higher than that of other options in the test results are extracted as the experience optimization decision set of the behavioral feature cluster; Establish a mapping relationship between the behavioral feature clusters of the specific scenic spot and the experience optimization decision set; integrate all mapping relationships of all specific scenic spots to construct the association rule base.

6. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S3, the specific process of initial trip planning is as follows: Parse the current user request to obtain a list of target attraction identifiers, and generate a set of attractions to be planned; match the set of attractions to be planned with the specific attraction set, and mark the identifiers belonging to the specific attraction set as nodes to be optimized; For each node to be optimized, query its mapped experience optimization decision set from the association rule base, and extract the attraction identifiers contained in the set; add the extracted attraction identifiers to the set of attractions to be planned to form an expanded attraction set; Based on the expanded set of attractions and the spatiotemporal constraint data between attractions, an initial attraction visit sequence is generated; the initial attraction visit sequence is traversed to identify the nodes to be optimized that are arranged as consecutive visit nodes; When consecutive nodes to be optimized are identified, the experience optimization decision set mapped to the first node to be optimized is obtained from the association rule base; Determine whether the second node to be optimized belongs to the scenic spot identifier in the experience optimization decision set; if it does not belong, select a scenic spot identifier from the experience optimization decision set and replace the second node to be optimized in the sequence or insert it before it; The adjusted attraction visit sequence is combined with traffic and time data to generate the initial itinerary plan.

7. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S4, the specific calculation process of the behavior feature vector is as follows: Obtain the real-time location sequence and dwell time sequence uploaded by the current user's mobile terminal; extract the departure time of the most recently visited attraction from the real-time location sequence, and extract the subsequent real-time location sequence with a fixed duration window starting from that time point; Based on the subsequent real-time positioning sequence within the fixed time window, calculate the movement speed sequence within the fixed time window, and calculate the rate of change of the movement speed sequence compared to the movement speed sequence within the control time window before the departure time point; Based on the dwell time sequence, the frequency of dwell events within the fixed duration window is statistically analyzed, and the rate of change of this frequency relative to the frequency of dwell events within the control duration window is calculated. The calculated rate of change of the movement speed sequence is combined with the rate of change of the dwell event frequency to generate the current user's current behavior feature vector.

8. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In S4, the specific construction process of the preset behavior offset classification model is as follows: From the association rule base, extract the cluster center feature vector of each behavioral feature cluster corresponding to all specific attractions; use each cluster center feature vector as a model training sample, and use the identifier of the behavioral feature cluster from which the cluster center feature vector originates as the category label of the training sample; Collect all training samples and category labels corresponding to all the specific attractions to construct a labeled training sample set; use the labeled training sample set to perform supervised training on a multi-class machine learning model with a preset basic structure; The multi-class machine learning model that has been trained and has reached the preset accuracy index is saved as the behavior offset classification model.

9. The artificial intelligence-based tourism itinerary planning method according to claim 1, characterized in that, In step S5, the specific process of dynamically adjusting the remaining travel plans of the current user is as follows: Obtain the offset category to which the current user's real-time behavior belongs, as output by the behavior offset classification model; and query the corresponding experience optimization decision set from the association rule base according to the offset category. Obtain the current user's real-time geographic location information and the sequence of attractions to be visited in the remaining itinerary plan; using the real-time geographic location as the starting point, the experience optimization decision set as the candidate destination constraint, and combining real-time traffic data and attraction opening time data, generate subsequent itinerary segments; The generated subsequent itinerary segments are merged and compared with the original attraction sequences that have not been executed in the remaining itinerary plan. If the attraction identifier in the original attraction sequence does not belong to the experience optimization decision set, then the attraction identifier in the subsequent itinerary segments that belongs to the experience optimization decision set is used to replace the attraction identifier that does not belong. The new itinerary sequence, after fusion and replacement processing, is combined with real-time updated traffic movement data and estimated travel time to obtain the adjusted dynamic itinerary plan.

Images