Tourism information release management method and system based on artificial intelligence and big data

Abstract

The invention provides a tourism information release management method and system based on artificial intelligence and big data. The method comprises the following steps: collecting user emotional states monitored in real time and classifying the user emotional states; recommending suitable tourism activities to the user according to the emotional state of the user; an AR scene is generated according to the personalized activity recommendation list, a user interacts by selecting the AR scene, and the AR scene is adjusted and optimized in real time according to real-time interaction; according to feedback data collected in real time in an AR scene, optimizing recommendation planning and experience contents; the recommendation strategy is dynamically adjusted in combination with historical preference data of the user, emotion feedback of the user and particularity of the tourism scene; and multi-level data integration, real-time resource optimization and dynamic scheduling are realized. According to the invention, real-time feedback and dynamic adjustment of tourism management are realized, so that the user satisfaction is greatly improved, and the requirements of modern tourists on individuation, interactivity and instantaneity are met.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and big data, and in particular to a tourism information publishing management method and system based on artificial intelligence and big data. Background Art

[0002] With the booming tourism industry, the application of intelligent technology has gradually become an important means to improve user experience. Existing tourism information release and management systems often rely on static data recommendations and lack dynamic responses to user emotions and immediate needs. These systems usually generate travel recommendations based on users' historical behavior data, ratings, and comments, but this approach has significant limitations. First, users' emotional fluctuations during travel (such as fatigue, stress, or excitement) have an important impact on their choices, but most recommendation systems fail to effectively capture these emotional changes, resulting in reduced relevance and applicability of recommended content. Second, current systems usually lack real-time interactive capabilities and are difficult to flexibly adjust based on users' immediate feedback and environmental changes, which affects users' experience when facing inappropriate activities or attractions.

[0003] In addition, the rise of social media has made users more willing to share and interact with others during travel, but existing systems often fail to effectively integrate these social elements, limiting communication and experience sharing among tourists. As a result, users often face information overload and poor experience during travel, affecting their overall satisfaction. Summary of the invention

[0004] The purpose of the present invention is to propose a tourism information publishing management method and system based on artificial intelligence and big data, which effectively overcomes the shortcomings of the prior art, not only improves the overall travel experience of users, but also promotes the sustainable development of the tourism industry. By achieving real-time feedback and dynamic adjustment, the present invention can greatly improve user satisfaction and meet the needs of modern tourists for personalization, interactivity and immediacy.

[0005] In order to achieve the above object, a first aspect of the present invention provides a tourism information publishing management method based on artificial intelligence and big data, the method comprising:

[0006] S1. Collect and classify the user’s emotional state monitored in real time, including:

[0007] S1.1. Use mobile applications and wearable devices to collect users’ physiological and social data in real time;

[0008] S1.2, preprocess the collected data and extract sentiment features;

[0009] S1.3, using the sentiment classification model based on support vector machine to calculate the sentiment score according to the extracted sentiment features;

[0010] S1.4, setting a threshold according to the calculated emotion score to classify the emotion state;

[0011] S2. Recommend suitable travel activities to the user based on the user's emotional state, including:

[0012] S2.1, constructing a database containing multiple tourism activities; wherein the data set includes activity ID, activity description, suitable emotion type and emotion suitability score;

[0013] S2.2, calculate the matching score of each activity according to the sentiment score, sort the matching scores, select the top k activities with the highest matching degree, and then filter out the highest ranking activities by setting the matching threshold;

[0014] S2.3, generating ranking results, and using the results as a personalized activity recommendation list for the user;

[0015] S2.4. Collect user feedback in real time, calculate user satisfaction scores based on user feedback, and adjust subsequent activity recommendation strategies;

[0016] S3. Generate an AR scene based on the personalized activity recommendation list. Users interact by selecting the AR scene, and adjust and optimize the AR scene in real time based on the real-time interaction, including:

[0017] S3.1. Generate AR scenes based on the personalized activity recommendation list, and dynamically adjust the AR scene content based on the real-time emotion score;

[0018] S3.2. Design a multimodal interaction method based on emotion, obtain users’ emotional feedback in real time during the user experience, and optimize the interactivity of the AR scene in real time based on this feedback. Finally, automatically record the user’s emotional state, interactive behavior, and feedback at the end of each user experience.

[0019] S4. Optimize recommendation planning and experience content based on real-time feedback data collected in AR scenarios, including:

[0020] S4.1. Construct a comprehensive feedback score function based on user feedback data combined with emotional state deviation and nonlinear adjustment factors to calculate the comprehensive feedback score;

[0021] S4.2. Adjust the priority of the next recommended activity list based on the comprehensive feedback score. If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted;

[0022] S4.3, adjusting the content complexity and interaction mode of the AR scene in real time according to the adjusted recommendation strategy, and monitoring the user's emotional state in real time for storage and analysis;

[0023] S5. Dynamically adjust the recommendation strategy based on the user's historical preference data, user's emotional feedback, and the particularity of the travel scenario, including:

[0024] S5.1, preprocess the user's historical preference data, comprehensive feedback score and current experience emotional state score, add exponential decay terms based on the preprocessed data to build a personalized priority adjustment model, and dynamically adjust the recommendation priority of each activity; wherein, the exponential decay term is used to measure the difference between the current user's emotional state and the emotional matching degree of the activity, and the priority is reduced as the difference increases.

[0025] S5.2. Sort the candidate activities according to the results output by the personalized priority adjustment model and select the top k best activities to generate a personalized recommendation list;

[0026] S5.3. Design an emotion fluctuation adjustment mechanism to deal with the situation where the user's emotional state fluctuates drastically, and adjust the recommendation strategy when the user's emotion fluctuates greatly;

[0027] S5.4. Set up an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation;

[0028] S6. Integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes personalized recommendation priority list, user feedback score, government resource data and merchant service data.

[0029] Furthermore, the sentiment score is calculated as follows:

[0030]

[0031] Among them, S represents the sentiment score, which is used to represent the user's emotional state; m represents the number of selected features; F i represents the value of the i-th feature; W i represents the weight of the i-th feature.

[0032] Furthermore, the matching score of each activity is calculated according to the sentiment score, which is expressed as follows:

[0033] MS i =EAS i ×S+λ@f(EAS i ,S)

[0034] Among them, MS i represents the matching score of the i-th activity; EAS i represents the emotional fitness score of the i-th activity; S represents the user's emotional score; λ represents the weight factor, which adjusts the impact of activity fitness on the final score; f(EAS i,S) represents an additional adaptation function that considers the nonlinear relationship between emotion and activity.

[0035] The following is a representation of the following method for calculating the user satisfaction score based on the user feedback and adjusting the subsequent activity recommendation strategy:

[0037]

[0038] Among them, FS represents the user feedback score; S i represents the user satisfaction score of the i-th activity; β represents the adjustment factor used to adjust the impact of recommendation feedback; g(RAL) represents the function based on the popularity of the recommended activity.

[0039] Furthermore, in the process of dynamically adjusting the AR scene content according to the emotion score, the emotion adaptation coefficient α is introduced to adjust the complexity of the scene content according to the user's emotion score and the scene adaptability, which is expressed as follows:

[0040]

[0041] Among them, AR adjusted represents the adjusted AR scene, which adapts to the user's current emotional state; α represents the emotional adaptation coefficient, which controls the sensitivity of the AR scene content to emotional changes; S represents the user's current emotional score, which represents the user's emotional feedback; μ S represents the mean of the global sentiment score; σ S Indicates the standard deviation of the sentiment score, used to standardize the sentiment value; EAS i represents the emotional fitness score of the ith activity, which is used to indicate how the scene matches the current emotional state;

[0042] The multimodality includes vision, audio and gesture; wherein the adjustment formula of the multimodal interaction capability is expressed as follows:

[0043] R AR =γ1@I v +γ2@I a +γ3·I g +λ@H(S)

[0044] Among them, R AR Represents the overall interactive response capability of the AR system; I v Represents the interactive strength of visual input; I a Indicates the interactive intensity of the audio input; I g represents the interaction strength of gesture input; γ1, γ2, and γ3 represent the weights used to control different input methods, which are adjusted based on the feedback of the user's current emotional state; H(S) represents the emotional state function, which reflects the nonlinear regulation of emotion on interactive operations.

[0045] Furthermore, the feedback comprehensive scoring function is expressed as follows:

[0046] CFS=w1·FS+w2·IF+w3·(S real -S prev ) 2

[0047] Among them, CFS represents the comprehensive feedback score, which reflects the user's overall feedback on the current experience; FS represents the user's subjective satisfaction score; IF represents the interaction frequency, which measures the depth of user participation; S real Indicates the emotional state score of the current experience; S prev Indicates the emotional state score of the previous experience; w1, w2, w3 represent weight parameters, which are used to adjust the influence of different feedback indicators;

[0048] The priority of the recommended activity list is adjusted based on the comprehensive feedback score, as shown below:

[0049]

[0050] Among them, P new (AID i ) indicates the activity AID i The optimized priority of P prev (AID i ) indicates the activity AID i Initial priority set; CFS thresh represents the feedback score threshold. When CFS is lower than this threshold, it indicates that the user experience is poor. β represents the adjustment factor, which is used to control the impact of feedback on the recommendation priority. H(AID i Represents a historical record function, which is used to calculate the user's past experience with the activity AID i λ represents the weight coefficient, which adjusts the proportion of historical preference in the priority.

[0051] Furthermore, if the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted, as shown below:

[0052]

[0053] Among them, R adjust represents the recommendation rhythm adjustment coefficient, which controls the frequency and intensity of activity recommendations; α represents the emotional state adjustment coefficient, which controls the impact of emotional fluctuations on the recommendation rhythm; γ represents the feedback adjustment coefficient, which controls the impact of feedback ratings on the recommendation rhythm; T S Indicates the regulatory time constant of the emotional state and determines the sensitivity of emotional fluctuations.

[0054] Furthermore, the content complexity and interaction mode of the AR scene are adjusted in real time according to the adjusted recommendation strategy, as shown below:

[0055]

[0056] Among them, AR opt Indicates the complexity of optimized AR content; AR prev represents the complexity of the original AR content; ρ represents the adjustment coefficient, which controls the degree of content simplification; L(AR) represents the redundancy loss function of the AR content, which is used to measure the interaction complexity of the current scene;

[0057] If the feedback score CFS is lower than the threshold feedback score threshold CFS thresh , simplifies AR content, reduces interaction frequency, reduces redundant information, and improves user comfort.

[0058] Furthermore, the recommendation priority calculation of the personalized priority adjustment model is expressed as follows:

[0059]

[0060] Among them, P opt (AID i ) represents the i-th activity AID i The optimized priority of H(AID i ) represents the user's historical preference weight, which represents the user's past preference for the activity AID i interest; CFS represents the user's comprehensive feedback score in the last experience; CFS max represents the highest feedback score of the user in history, which is used to normalize the current feedback score; α, β, γ respectively control the influence of historical preference, feedback score and emotional state on recommendation priority; S real Indicates the user’s current emotional state; S target (AID i ) indicates the activity AID i The most suitable emotional state;

[0061] The exponential decay term is exp(-|S real -S target (AID i )|), which is used to measure the difference between the current user's emotional state and the activity's emotional match, and quickly reduce the priority as the difference increases;

[0062] The emotional fluctuation regulation mechanism is expressed as follows:

[0063]

[0064] Among them, Padjust (AID i ) represents the adjusted activity priority, reflecting the change in priority after emotional fluctuations; θ represents the adjustment coefficient, controlling the impact of emotional fluctuations on priority; |S real -S target (AID i )| represents the difference between the user’s current emotional state and the target emotional state of the recommended activity.

[0065] Furthermore, the S6 specifically includes:

[0066] S6.1. Set up a dynamic resource matching model based on user demand and resource supply to dynamically match inbound and outbound resources during peak tourism periods;

[0067] S6.2, collect user feedback in real time, adjust resource matching according to the deviation of each feedback, and design traffic and passenger flow optimization models to intelligently dispatch scenic spots by combining real-time traffic data and passenger flow prediction information;

[0068] S6.3. Upload real-time multi-level data to the hierarchical data interaction platform, and interact with local governments, businesses and user devices in real time through APIs to ensure that all parties can obtain the latest resource status and user feedback in a timely manner and make corresponding scheduling adjustments;

[0069] S6.4. Record and analyze feedback data and resource allocation history to identify bottlenecks in long-term resource allocation and optimize resource utilization efficiency.

[0070] In a second aspect of the present invention, a tourism information publishing management system based on artificial intelligence and big data is provided, the system comprising:

[0071] The emotion classification subsystem is used to collect and classify the user's emotional state monitored in real time, and performs the following steps:

[0072] S1.1. Use mobile applications and wearable devices to collect users’ physiological and social data in real time;

[0073] S1.2, preprocess the collected data and extract sentiment features;

[0074] S1.3, using the sentiment classification model based on support vector machine to calculate the sentiment score according to the extracted sentiment features;

[0075] S1.4, setting a threshold according to the calculated emotion score to classify the emotion state;

[0076] The personalized recommendation subsystem is used to recommend suitable travel activities to users based on their emotional state. It performs the following steps:

[0077] S2.1, constructing a database containing multiple tourism activities; wherein the data set includes activity ID, activity description, suitable emotion type and emotion suitability score;

[0078] S2.2, calculate the matching score of each activity according to the sentiment score, sort the matching scores, select the top k activities with the highest matching degree, and then filter out the highest ranking activities by setting the matching threshold;

[0079] S2.3, generating ranking results, and using the results as a personalized activity recommendation list for the user;

[0080] S2.4. Collect user feedback in real time, calculate user satisfaction scores based on user feedback, and adjust subsequent activity recommendation strategies;

[0081] The AR interaction subsystem is used to generate AR scenes based on the personalized activity recommendation list. Users interact by selecting AR scenes, and the AR scenes are adjusted and optimized in real time based on real-time interactions. The following steps are performed:

[0082] S3.1. Generate AR scenes based on the personalized activity recommendation list, and dynamically adjust the AR scene content based on the real-time emotion score;

[0083] S3.2. Design a multimodal interaction method based on emotion, obtain users’ emotional feedback in real time during the user experience, and optimize the interactivity of the AR scene in real time based on this feedback. Finally, automatically record the user’s emotional state, interactive behavior, and feedback at the end of each user experience.

[0084] The AR scene optimization subsystem is used to optimize the recommendation planning and experience content based on the real-time feedback data collected in the AR scene. The following steps are performed:

[0085] S4.1. Construct a comprehensive feedback score function based on user feedback data combined with emotional state deviation and nonlinear adjustment factors to calculate the comprehensive feedback score;

[0086] S4.2. Adjust the priority of the next recommended activity list based on the comprehensive feedback score. If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted;

[0087] S4.3, adjusting the content complexity and interaction mode of the AR scene in real time according to the adjusted recommendation strategy, and monitoring the user's emotional state in real time for storage and analysis;

[0088] The activity optimization subsystem is used to dynamically adjust the recommendation strategy by combining the user's historical preference data, the user's emotional feedback, and the particularity of the travel scene. It performs the following steps:

[0089] S5.1, preprocess the user's historical preference data, comprehensive feedback score and current experience emotional state score, add exponential decay terms based on the preprocessed data to build a personalized priority adjustment model, and dynamically adjust the recommendation priority of each activity; wherein, the exponential decay term is used to measure the difference between the current user's emotional state and the emotional matching degree of the activity, and the priority is reduced as the difference increases.

[0090] S5.2. Sort the candidate activities according to the results output by the personalized priority adjustment model and select the top k best activities to generate a personalized recommendation list;

[0091] S5.3. Design an emotion fluctuation adjustment mechanism to deal with the situation where the user's emotional state fluctuates drastically, and adjust the recommendation strategy when the user's emotion fluctuates greatly;

[0092] S5.4. Set up an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation;

[0093] The system optimization subsystem is used to integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes personalized recommendation priority lists, user feedback scores, government resource data and merchant service data.

[0094] The beneficial technical effects of the present invention are at least as follows:

[0095] The present invention proposes an intelligent tourism experience optimization system based on emotional computing and augmented reality technology. Its main innovation is to monitor the user's emotional state in real time through emotional analysis, dynamically adjust travel recommendations, and ensure that activities and attractions that match the user's current feelings are provided. This system can not only respond to the user's emotional fluctuations in a timely manner, but also combine augmented reality technology to provide immersive cultural experience and social interaction, enhancing the user's sense of participation and interactivity. At the same time, the system integrates a multi-level data interaction platform, so that users, tourism businesses and local governments can communicate efficiently, realize information sharing and optimize resource allocation. This comprehensive solution effectively overcomes the shortcomings of the existing technology, not only improves the user's overall travel experience, but also promotes the sustainable development of the tourism industry. By realizing real-time feedback and dynamic adjustment, the present invention can greatly improve user satisfaction and meet the needs of modern tourists for personalization, interactivity and immediacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0097] Figure 1This is a flow chart of the tourism information publishing management method based on artificial intelligence and big data of the present invention. DETAILED DESCRIPTION

[0098] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0099] like Figure 1 As shown, the tourism information publishing management method based on artificial intelligence and big data provided by the embodiment of the present invention includes the following steps S1-S6:

[0100] S1. Collect and classify the user's emotional state in real time.

[0101] Specifically, mobile applications and wearable devices (such as smart watches) are used to collect users’ physiological and social data in real time, including:

[0102] Physiological data: Heart rate (HR): The heart rate sensor collects the user's heartbeats per minute. Step frequency (SF): The accelerometer monitors the user's steps and calculates the number of steps per minute. Skin temperature (ST): The temperature sensor records the user's skin temperature (unit: ℃) every minute.

[0103] Social media activities: Status updates posted by users (SU): Extract recent status updates from users’ social media accounts through APIs and analyze sentiment. Emotion tags (ET): Perform sentiment analysis on status updates through natural language processing technology and generate sentiment tags (such as positive, negative, neutral).

[0104] These data are transmitted to the central database in real time through the API interface for subsequent analysis and processing.

[0105] Furthermore, in the collected raw data, data preprocessing is first performed, including denoising, normalization, etc. Then, feature weights are calculated using the following formula to ensure that the contribution of each feature to sentiment analysis is reasonably evaluated:

[0106]

[0107] Among them, W i V represents the weight of the i-th feature. i Represents the original value of the i-th feature, such as the value of heart rate, cadence, and skin temperature. n represents the number of features.

[0108] For example, if the user's heart rate is 80bpm, the cadence is 100 steps / min, and the skin temperature is 36.5℃, the weight of each feature is calculated. Through normalization, this ensures that the influence of each feature in the sentiment analysis matches its importance.

[0109] Furthermore, next, a sentiment classification model based on support vector machine (SVM) is used to calculate sentiment scores using the extracted features. The specific formula is as follows:

[0110]

[0111] Where S is the emotion score, which is used to represent the user's emotional state. m is the number of selected features (e.g., heart rate, cadence, skin temperature). i Represents the value of the i-th feature.

[0112] Assuming that the heart rate weight is 0.4, the step frequency weight is 0.3, the skin temperature weight is 0.3, and the corresponding feature values ​​are 80, 100, and 36.5, the emotion score is:

[0113] S=(0.4·80)+(0.3·100)+(0.3·36.5)=32+30+10.95=72.95

[0114] Furthermore, according to the calculated emotion score S, a threshold is set to classify the emotion state. To set the emotion threshold T, the specific steps are as follows:

[0115] If S>T1 (for example, set T1=70), the user state is "happy".

[0116] If T1 ≥ S > T2 (set T2 = 40), the user status is "neutral".

[0117] If S≤T2, the user status is "passive".

[0118] This classification enables the system to determine the emotional state of a user in real time every time his / her data is updated, and use this information as input for the subsequent step (emotion-activity matching).

[0119] S2. Recommend suitable travel activities to users based on their emotional state.

[0120] Firstly, a database containing a variety of tourism activities is established. Each activity record in the database includes the following information: activity ID (AID), activity description (AD), suitable emotion type (ET) (such as "pleasure", "neutral", "negative"), and emotion adaptation score (EAS), which indicates the matching degree of the activity with different emotional states.

[0121] In the database, the suitability score of activities is constructed based on historical data and user feedback to ensure the accuracy of activity recommendations. For example, the activity "Beach Relaxation" may have the following records:

[0122] AID1=1

[0123] AD1 = "Relax on the beach"

[0124] ET1 = "pleasure"

[0125] EAS1=0.9

[0126] Further, according to the sentiment score S in step 1 (e.g., S=72.95), the matching score (MS) of each activity is calculated using the following formula:

[0127] MS i =EAS i ×S+λ·f(EAS i ,S)

[0128] Among them, MS i Represents the matching score of the i-th activity. i represents the emotional fitness score of the ith activity. S represents the user's emotional score. λ represents the weight factor, which adjusts the impact of activity fitness on the final score. f(EAS i ,S) represents an additional adaptation function, considering the nonlinear relationship between emotion and activity, for example: This function can enhance the recommendation weight of highly suitable activities when the user's emotion is low. This means that if the user's emotional state is low, the system will be more inclined to recommend highly suitable activities.

[0129] Furthermore, the matching score MS of all activities is i Sort and select the top k activities with the highest matching degree. By setting the threshold T MS , filter out suitable activities:

[0130] T MS =min(MS)

[0131] When MS i >T MS When , the activity will be included in the recommendation list. Assuming that the present invention calculates the MS from k activities to be {65.655, 70.5, 68.3}, the first three activities are finally retained.

[0132] Furthermore, based on the sorting results, a personalized activity recommendation list (RAL) for the user is generated, and the structure is as follows:

[0133] RAL={(AID1,AD1),(AID2,AD2),(AID3,AD3)}

[0134] For example, the recommended list is:

[0135] RAL = {(1,"Beach relaxation"),(2,"Cultural visits"),(3,"Hiking")}

[0136] This list will ensure that users can choose the activity that best matches their emotional state.

[0137] Furthermore, a real-time feedback mechanism is introduced. After the user completes the recommended activity, the system will adjust the subsequent activity recommendation strategy based on the user's satisfaction score (FS). The feedback score calculation formula is as follows:

[0138]

[0139] Where FS represents the user feedback score. i represents the user satisfaction score of the i-th activity. β represents the adjustment factor used to adjust the impact of recommendation feedback. g(RAL) represents a function based on the popularity of the recommended activity, for example: This function can enhance the influence of users on the feedback of recommendation activities.

[0140] Furthermore, based on real-time feedback and historical data, the system will dynamically optimize the recommendation strategy to ensure that subsequent activity recommendations are more in line with the user's emotional changes. For example, if the user gives a high satisfaction score after "relaxing on the beach", the system will increase the proportion of such activities in future recommendations.

[0141] S3. Generate an AR scene based on the personalized activity recommendation list. Users interact by selecting the AR scene, and the AR scene is adjusted and optimized in real time based on real-time interaction.

[0142] Specifically, from the output of step 2 - the recommended activity list RAL (which includes the IDAID of each activity i and describe AD i ), select AR scenes corresponding to these activities. The AR scene library is pre-built according to different activity types (such as "cultural visit", "hiking", etc.), covering a wealth of virtual information, interactive elements and immersive experience. The scene selection formula is as follows:

[0143] ARS i ={AID i ,SID i}

[0144] Among them, ARS i Represents the AR scene related to the i-th recommended activity.i Indicates the ID of the i-th recommended activity. SID i Indicates the AR scene ID that matches the activity ID. For example, if the recommended activity AID1 is "Beach Relaxation", the corresponding AR scene SID1 can be "Virtual Beach Scene", including relaxing music, landscape simulation, etc. SID1 is the index of the virtual scene, which is used as the specific AR module called in the system.

[0145] Furthermore, in the generated AR scene, the system dynamically adjusts the AR content according to the user's emotion score S (the real-time score obtained from emotion monitoring in step 1). This adjustment not only changes the interaction frequency in the scene, but also adaptively enhances or reduces elements in the scene, such as prompt information, virtual tour guide explanations, etc., based on the user's real-time emotional state.

[0146] Furthermore, an emotion adaptation coefficient α is introduced, which adjusts the complexity of AR content according to the user's emotion score S and scene adaptability EAS. The formula is as follows:

[0147]

[0148] Among them, AR adjusted represents the adjusted AR scene, which adapts to the user's current emotional state. α represents the emotional adaptation coefficient, which controls the sensitivity of the AR scene content to emotional changes. S represents the user's current emotional score, which is obtained in real time from step 1. μ S Represents the mean of the global sentiment score. S Indicates the standard deviation of the sentiment score, used to standardize the sentiment value. EAS i represents the emotional fitness score of the ith activity (obtained from step 2), which is used to indicate how the scene matches the current emotional state. For example, if the user's emotional score is low (S<μ S ), the system will increase the user's comfort by reducing distracting elements in the scene (such as excessive prompt information or complex interactions). At the same time, if S is high (S>μ S ), the system may increase interactivity, such as providing more virtual scenery, explanations, or social interaction opportunities.

[0149] Furthermore, the system provides multimodal interaction for each AR scene, including visual, voice, gesture and other input methods. This interaction method needs to respond to the user's emotional state in real time and enhance or simplify the interaction operation in a timely manner according to the user's emotional feedback. The adjustment formula of multimodal interaction capability is as follows:

[0150] R AR =γ1·I v +γ2·I a +γ3·I g +λ·H(S)

[0151] Among them, R AR Represents the overall interactive response capability of the AR system. v Indicates the interactive intensity of visual input (such as users controlling virtual objects through gaze and sight). a Indicates the interactive intensity of audio input (e.g., users interact with the virtual tour guide through voice commands). g Represents the interactive strength of gesture input (e.g., users operate virtual objects through gestures). γ1,γ2,γ3 represent the weights used to control different input methods, which are adjusted based on the feedback of the user's current emotional state. H(S) represents the emotional state function, which reflects the nonlinear adjustment of emotion on interactive operations.

[0152] For example, when the user's emotional state is positive (S is higher), γ3 (gesture interaction weight) can be increased because the user may be willing to perform more interactive operations; while when the user's emotional state is low, γ1 (visual interaction) and H(S) will increase to reduce the complexity of the user's gesture interaction.

[0153] Furthermore, during the user experience, the system continuously obtains the user's emotional feedback (such as satisfaction score, interaction frequency, etc.) through sensors, and optimizes the interactivity of the AR scene in real time based on this feedback. For example, if the user's emotional score drops during the "cultural tour" experience, the system will dynamically reduce the amount of information and interaction frequency in the scene to ensure that the user's emotional state will not be further reduced due to information overload. The real-time adjustment formula is as follows:

[0154]

[0155] Among them, AR final Indicates an AR scene that has been optimized in real time. FS indicates the user's real-time feedback score (derived based on user satisfaction during interaction). thresh It represents the feedback threshold. When FS is lower than this threshold, the system will reduce the complexity in the AR scene. β represents the adjustment coefficient, which controls the sensitivity of feedback to AR content adjustment.

[0156] For example, if FS is significantly lower than F thresh , then AR final The interactivity and information presentation intensity will be significantly reduced to avoid further deterioration of the user's emotional state.

[0157] Furthermore, the system automatically records the user's emotional state, interactive behavior, and feedback at the end of each user experience. These records will be used to optimize future AR interaction recommendations. For example, the system can identify the user's preference for certain virtual scenes or interaction methods and enhance these preference elements in future recommendations. This process is optimized by the following formula to update the weight:

[0158] W opt =W prev +η·(FS-W prev )

[0159] Among them, W opt Represents the optimized weight (such as the user's preference weight for a certain interaction method). prev Represents the last weight record. η represents the learning rate, which determines the speed of system optimization. FS represents the user feedback score.

[0160] Through this process, the system will continuously update the user's preferences and apply them to subsequent activity recommendations and AR scene construction, ensuring that each experience is more personalized and efficient.

[0161] Through the above steps, the development of augmented reality interactive experience not only adaptively adjusts according to the user's emotional state, but also ensures the user's immersion and participation through multimodal interaction, real-time feedback and personalized optimization.

[0162] S4. Optimize recommendation planning and experience content based on real-time feedback data collected in AR scenarios.

[0163] Specifically, the system first collects user feedback data in real time from multi-source sensors and interaction records in the AR experience. Key data includes:

[0164] Emotional state feedback (S real ): The user's emotional state is obtained through physiological and behavioral signals such as heart rate, facial expressions, and steps, as a real-time expression of the user's actual emotional state.

[0165] Satisfaction score (FS): The user's subjective evaluation of the current experience, which can be collected through questionnaires, voice or gesture interactions, or inferred through algorithms combined with emotional state feedback.

[0166] Interaction frequency (IF): The number of times a user interacts in an AR scene, such as touching virtual objects and using voice commands, which reflects the user's engagement and experience.

[0167] The real-time feedback data will be used as input for this step and for the next step of feedback scoring and optimization strategy calculation.

[0168] Furthermore, multiple dimensions of real-time feedback (emotional state, satisfaction score, interaction frequency, etc.) will be integrated into a comprehensive feedback score (CFS) to evaluate the effect of the current AR scene and provide a reference for subsequent recommendation optimization. This invention introduces an improved weighted feedback model, combining emotional state deviation and nonlinear adjustment factors to make the comprehensive score more responsive. The formula is as follows:

[0169] CFS=w1·FS+w2·IF+w3·(S real -S prev ) 2

[0170] Among them, CFS represents the comprehensive feedback score, which reflects the user's overall feedback on the current experience. FS represents the user's subjective satisfaction score. IF represents the interaction frequency, which measures the depth of user participation. real Indicates the emotional state score of the current experience. prev represents the emotional state score of the previous experience. w1, w2, w3 represent weight parameters, which are used to adjust the influence of different feedback indicators. real -S prev ) 2 It is the quadratic term of the change in emotional state, reflecting the impact of the fluctuation of user emotions on the feedback score. If the emotional state fluctuates greatly (whether positive or negative), the system will amplify it to indicate that the current scene may have a greater impact on emotions. Through this innovative design, the system can be more sensitive to changes in emotional state so as to adjust subsequent recommendations in a timely manner.

[0171] Furthermore, after obtaining the comprehensive feedback score CFS, the system needs to adjust the priority of the next recommended activity list (RAL) based on the score. This adjustment is not only based on the user's current feedback, but also combined with the user's preference records in the previous steps. The optimized activity priority calculation formula is as follows:

[0172]

[0173] Among them, P new (AID i ) indicates the activity AID i The optimized priority of P prev (AID i ) indicates the activity AID i The initial priority set in step 2. CFS thresh It represents the feedback score threshold. When CFS is lower than this threshold, it indicates that the user experience is poor. β represents the adjustment factor, which is used to control the impact of feedback on the recommendation priority. i ) represents the history function, which is used to calculate the user's past experience with the activity AID i λ represents the weight coefficient, which adjusts the proportion of historical preference in the priority. In this formula, in addition to the impact of the comprehensive feedback score on the priority, an additional preference term is introduced This item is based on the user's past preference for similar activities H (AID i), to prevent the user's historical preferences from being completely overturned based on just one feedback. This two-way optimization mechanism not only maintains the continuity of the user experience, but also can flexibly respond to changes in real-time feedback.

[0174] Furthermore, based on the changing trend of the emotional state, the system can also adjust the type and rhythm of recommended activities in real time. If it detects a large fluctuation in the user's emotional state (for example, a sudden change from "happy" to "negative"), the system will automatically adjust the recommendation strategy to avoid further stimulating the user's emotional fluctuations.

[0175] The recommended rhythm adjustment formula is as follows:

[0176]

[0177] Among them, R adjust represents the recommendation rhythm adjustment coefficient, which controls the frequency and intensity of activity recommendations. α represents the emotional state adjustment coefficient, which controls the impact of emotional fluctuations on the recommendation rhythm. γ represents the feedback adjustment coefficient, which controls the impact of feedback ratings on the recommendation rhythm. S Indicates the regulatory time constant of the emotional state and determines the sensitivity of emotional fluctuations.

[0178] If the user's emotions fluctuate too much, the system will reduce the frequency of recommended activities or recommend some low-stimulation activities (such as relaxing activities) to avoid further deterioration of the user's emotional state. At the same time, if the feedback score CFS is low, the system will make more frequent recommendations to enhance the user's positive experience.

[0179] Furthermore, in addition to optimizing recommended activities, the system also needs to adjust the content complexity and interaction mode of the AR experience based on real-time feedback. The specific content optimization introduces an innovative regularization term to control the redundancy of interactive content to reduce the burden on users when their emotional state is poor. The content adjustment formula is as follows:

[0180]

[0181] Among them, AR opt Indicates the complexity of the optimized AR content. prev Represents the complexity of the original AR content. ρ represents the adjustment coefficient, which controls the strength of content simplification. L(AR) represents the redundancy loss function of the AR content, which is used to measure the interaction complexity of the current scene.

[0182] If CFS is far below the threshold CFS thresh , the system will simplify AR content, reduce the frequency of interaction, and reduce redundant information to improve user comfort.

[0183] Furthermore, after the system completes the optimization strategy after each experience feedback, it will store and analyze the user's emotional state, activity preferences and AR interaction records for a long time. Each user's data will be used to gradually build a personalized model, thereby improving the accuracy of the system's recommendations in future experiences. The update of the personalized model uses a gradient optimization algorithm:

[0184] W opt =W prev +η·(CFS-W prev )

[0185] Among them, W opt represents the updated personalized weight. prev represents the last personalized weight. η represents the learning rate, which determines the speed of model update. Through this mechanism, the system can gradually improve its understanding of the user's emotional state, preferences, and interaction habits, ensuring that future recommendations and AR experiences are more in line with the user's personalized needs.

[0186] S5. Dynamically adjust the recommendation strategy based on user historical preference data, user emotional feedback, and the particularity of the travel scenario.

[0187] It can be understood that the input data of this step comes from the output of the real-time feedback and recommendation optimization in the previous step, including:

[0188] Comprehensive Feedback Score (CFS): indicates the user's overall feedback score for the previous AR experience, which has been calculated in the previous step.

[0189] Historical preference weight (H(AID i )): User historical preferences, which represent the user’s past interaction preferences and emotional state records with specific types of activities.

[0190] Real-time emotional state (S real ): The user’s current emotional state monitoring data obtained from step 1 is used to measure the fit between the current emotion and the recommended activity.

[0191] These data must first be standardized to eliminate the dimensional differences of different variables. The standardization formula is as follows:

[0192]

[0193] in, Indicates standardized data (such as CFS, H(AID i ) etc.). X represents the original input data. μ X Represents the mean of the data. X Represents the standard deviation of the data.

[0194] Through standardization, the uniformity of subsequent calculations is ensured, so that data from different sources can be analyzed and optimized at the same scale.

[0195] Furthermore, in order to meet the complexity of tourism scenarios, the present invention designs a personalized priority adjustment model, which combines the user's real-time feedback, historical preferences and current emotional state to dynamically adjust the recommendation priority of each activity. The core formula of the recommendation priority is as follows:

[0196]

[0197] Among them, P opt (AID i ) represents the i-th activity AID i The optimized priority of H(AID) is used for sorting and recommendation. i ) represents the user's historical preference weight, which represents the user's past preference for the activity AID i CFS represents the comprehensive feedback score of the user in the last experience, which comes from real-time feedback and recommendation optimization steps. max S represents the highest feedback score of the user in history, which is used to normalize the current feedback score. α, β, and γ respectively control the influence of historical preference, feedback score, and emotional state on recommendation priority. real Indicates the user’s current emotional state, provided by the real-time emotion monitoring module. target (AID i ) indicates the activity AID i The most suitable emotional state.

[0198] Furthermore, the core of this formula is to add the exponential decay term exp(-|S real -S target (AID i )|), which is used to measure the difference between the current user's emotional state and the emotional match of the activity, and quickly reduce the priority as the difference increases. This innovation effectively solves the problem that the recommendation system in the existing technology fails to fully consider the fluctuation of user emotions, ensuring that the recommended activities are highly matched with the user's emotions.

[0199] Furthermore, based on the results calculated by the above priority model, the system will sort all candidate activities and select the top k best activities to generate a personalized recommendation list. The sorting rules are as follows:

[0200] RAL opt ={AID1, AID2, …, AID k ∣P opt (AID1)>P opt (AID2)>…>P opt (AID k)}

[0201] Recommended activities list RAL opt This will directly affect the subsequent recommended content. In the tourism scenario, this dynamic sorting mechanism ensures that the system can respond to changes in the user's emotional state in real time, while taking into account the user's long-term interest preferences, so that each recommended activity can effectively meet the user's needs.

[0202] Furthermore, during the travel process, the user's emotional state may fluctuate dramatically (for example, suddenly changing from "happy" to "negative"). In order to deal with such emotional fluctuations, this solution designs an emotional fluctuation adjustment mechanism to adjust the recommendation strategy when the user's emotional fluctuations are large, reduce the recommendation frequency of highly stimulating activities, and increase the frequency of relaxing activities. The adjustment formula is as follows:

[0203]

[0204] Among them, P adjust (AID i ) represents the adjusted activity priority, reflecting the change in priority after emotional fluctuations. θ represents the adjustment coefficient, controlling the impact of emotional fluctuations on priority. |S real -S target (AID i )| represents the difference between the user’s current emotional state and the target emotional state of the recommended activity.

[0205] Through this formula, the system can automatically adjust the rhythm of recommended content. For example, when the user's emotions fluctuate violently, the system will prioritize recommending some activities with higher emotional matching to reduce the negative impact of the user's emotional state and improve the user's emotional recovery speed.

[0206] Furthermore, in order to further improve the accuracy of personalized recommendations, the system also designs an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation. The present invention introduces a learning rate adjustment formula to enable the system to speed up learning when user feedback is good and slow down the adjustment pace when feedback is poor, thus avoiding the recommendation inaccuracy problem caused by excessive adjustment. The learning rate update formula is as follows:

[0207]

[0208] Among them, η new Represents the new learning rate, which is used to update the learning ability of the model. prev Indicates the last learning rate. κ indicates the learning rate adjustment coefficient. CFS indicates the current comprehensive feedback score. avg Indicates the average of historical feedback scores. avg, the system will speed up the learning process to quickly adapt to changes in user preferences; on the contrary, when CFS <CFS avg , the learning rate is reduced to avoid overfitting.

[0209] Furthermore, in the end, the system combines the user's long-term preferences with real-time feedback, gradually adjusts the weight of each recommendation activity through multiple feedback records, and establishes a personalized model for the user. The result of each recommendation adjustment will be stored in the user's preference profile for reference in subsequent recommendation strategies. The core formula of this mechanism is:

[0210] W opt (AID i )=W prev (AID i )+η new ·(P adjust (AID i )-W prev (AID i ))

[0211] Among them, W opt (AID i ) indicates the activity AID i The final personalized weight W prev (AID i ) represents the activity weight of the last record. adjust (AID i ) represents the weight after adjustment by emotional fluctuation and priority. new Represents the learning rate, which is used to control the magnitude of weight adjustment.

[0212] By integrating long-term preferences with short-term feedback, the system can accurately capture changes in user preferences and continuously optimize recommendation strategies to ensure that each recommended content is highly personalized and emotionally adapted.

[0213] The personalized recommendation adjustment algorithm effectively improves the accuracy of recommended content by combining the user's historical preferences, real-time feedback and emotional fluctuations. Through the emotional fluctuation adjustment mechanism and adaptive learning strategy, the system can continuously optimize the recommendation strategy to ensure that the user's emotional state matches the recommended content, thereby achieving personalized improvement of the travel experience.

[0214] S6. Integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes personalized recommendation priority list, user feedback score, government resource data and merchant service data.

[0215] Specifically, the input data of this step mainly includes:

[0216] Personalized Recommendation Priority List RAL opt: This list comes from the personalized recommendation adjustment algorithm in the fifth step and represents the recommendation priority of each activity.

[0217] User Feedback Score (CFS): The user's comprehensive feedback score on the previous activity or experience, which is used to adjust subsequent recommendations.

[0218] Government Resource DataG r (t): Resource data provided by local governments, such as scenic spot opening hours, site capacity, real-time traffic flow and other information.

[0219] Merchant Service Data b (t): Service resources provided by merchants, such as real-time supply information of catering, accommodation and entertainment activities.

[0220] These data will be integrated through a multi-level data interaction platform and processed based on a unified data format. In order to ensure the consistency of data processing, the input data needs to be normalized to unify the scales of different data sources. The normalization formula is as follows:

[0221]

[0222] Among them, X norm Represents the standardized data value, covering G r (t) and S b (t) and other variables. min(X) and max(X) represent the minimum and maximum values ​​in the input data, ensuring that the data is in the range [0,1].

[0223] Through this data standardization process, the scale differences between different data sources can be effectively eliminated, allowing them to be seamlessly integrated and processed in subsequent optimization models.

[0224] Furthermore, the resources in the tourism industry are often limited, especially during peak periods, when tourism resources such as transportation, scenic spot capacity, and merchant services are in short supply. This step proposes a resource dynamic matching model based on user demand and resource supply to ensure that user recommended activities can be effectively matched with resource conditions. Define an innovative resource supply and demand matching formula:

[0225]

[0226] Among them, R opt (AID i ) represents the resource matching optimization index of the ith activity, indicating the degree of support of the current resources for this activity. RAL opt (AID i ) represents the user recommendation priority of the i-th activity (output from the fifth step). rS (t) represents the availability of government resources, which measures the resources of the scenic spot at a specific time point t (such as opening hours, venue capacity, etc.). b (t) indicates the merchant service availability, reflecting the AID of the activity at a specific time point t i The supply of related services.

[0227] This formula balances user demand (reflected by personalized recommendation priorities) with resource supply. When government resources or merchant services for an activity become tight, the system automatically lowers the recommendation weight for that activity to ensure that users do not experience a poor experience when resources are tight. Conversely, when resources for an activity are more plentiful, the system prioritizes that activity to balance resource usage and optimize user experience.

[0228] Furthermore, in order to cope with the dynamic changes in user feedback and tourism resource status, the system needs to be adjusted through a real-time feedback mechanism. Based on the user's real-time feedback score CFS (provided by the previous step), combined with real-time government and merchant data, the system can adjust the resource matching degree of recommended activities according to the user's actual experience. The formula for feedback interaction and resource adjustment is designed as follows:

[0229]

[0230] Among them, R adjust (AID i ) represents the resource matching index after feedback adjustment. λ represents the adjustment coefficient, which indicates the sensitivity of feedback to recommendation adjustment. ΔCFS represents the difference between the current feedback score and the last feedback score, which is used to measure the fluctuation of user experience. CFS max Indicates the highest feedback score in history.

[0231] This formula reflects the dynamic impact of user feedback on resource scheduling. For example, if the user's feedback score for an activity drops significantly (|ΔCFS| is large), the system will automatically reduce the priority and resource allocation of the activity to avoid recommending activities with poor experience. Conversely, if the user's feedback score for an activity is high, the system will prioritize allocating more resources to meet user needs and improve user satisfaction.

[0232] Furthermore, during the travel process, traffic and crowd management are one of the core challenges. Especially during peak hours, the traffic scheduling and crowd management of tourists will greatly affect the user's travel experience. To this end, the present invention designs a traffic and crowd optimization model, which realizes the intelligent scheduling of scenic spots by combining real-time traffic data and crowd prediction information. The traffic and crowd optimization formula is as follows:

[0233]

[0234] Among them, Topt (t) represents the traffic optimization coefficient at time t, which is used to reflect the dynamic scheduling of traffic resources. adjust (AID i ) represents the resource matching index of the ith activity (after feedback adjustment). i ,t represents the active AID at time t i The tourist flow forecast is based on historical data and real-time flow calculation. capacity (t) represents the traffic capacity at time t (such as the carrying capacity of public transportation vehicles and routes, etc.). Through this formula, the system can automatically allocate traffic resources to ensure that the traffic carrying capacity is balanced with the flow of tourists in the scenic area. For example, in scenic areas with dense tourists, the system will increase the input of traffic resources to reduce congestion; while in relatively idle scenic areas, the system will reduce traffic dispatch to save resources.

[0235] Furthermore, the core of this step is to upload multi-level data such as user feedback, resource supply and demand, traffic flow, etc. to the multi-level data interaction platform in real time, and interact with local governments, businesses and user devices in real time through API. Through this platform, the system can realize efficient sharing and linkage of multi-party information, ensuring that all parties can obtain the latest resource status and user feedback in a timely manner, and make corresponding scheduling adjustments. The intelligent update formula for multi-level data interaction is as follows:

[0236] D update (t) = f(G r (t),S b (t),P adjust (AID i ),L(AID i ,t),T capacity (t))

[0237] Among them, D update (t) represents the amount of multi-level data update at time t. f(·) represents the data update function, which comprehensively considers government resources, merchant services, user feedback, traffic forecast and traffic capacity. Through this formula, the system will automatically adjust resource allocation according to the multi-level data at time t. Government departments can update (t) Optimize the opening hours and capacity limits of scenic spots; merchants can adjust service supply, such as increasing or decreasing the number of operations of certain tourism projects; users can also obtain more real-time activity recommendations and traffic information through mobile terminals to avoid resource shortages or congestion.

[0238] Furthermore, in order to ensure the long-term optimization of the system and the sustainability of resource utilization, the system records and analyzes all feedback data and resource allocation history. By mining these historical data, the system can identify bottlenecks in long-term resource allocation and gradually optimize resource utilization efficiency, improve user experience consistency and reasonable allocation of tourism resources. The system's long-term feedback optimization formula is as follows:

[0239] W opt (t+1)=W prev (t)+η·(D update (t)-W prev (t))

[0240] Among them, W opt (t+1) represents the resource allocation weight at time t+1. prev (t) represents the resource allocation weight at time t. η represents the learning rate, which is used to adjust the optimization speed of the system in long-term feedback. This formula is obtained by gradually adjusting W opt (t+1), ensuring that the system can continuously optimize resource allocation and ultimately provide users with a continuously improved travel experience.

[0241] In another embodiment of the present invention, a tourism information publishing management system based on artificial intelligence and big data includes:

[0242] A tourism information publishing and management system based on artificial intelligence and big data, the system comprising:

[0243] The emotion classification subsystem 301 is used to collect and classify the user's emotion state monitored in real time, and performs the following steps:

[0244] S1.1. Use mobile applications and wearable devices to collect users’ physiological and social data in real time;

[0245] S1.2, preprocess the collected data and extract sentiment features;

[0246] S1.3, using the sentiment classification model based on support vector machine to calculate the sentiment score according to the extracted sentiment features;

[0247] S1.4, setting a threshold according to the calculated emotion score to classify the emotion state;

[0248] The personalized recommendation subsystem 302 is used to recommend suitable travel activities to the user according to the user's emotional state, and performs the following steps:

[0249] S2.1, constructing a database containing multiple tourism activities; wherein the data set includes activity ID, activity description, suitable emotion type and emotion suitability score;

[0250] S2.2, calculate the matching score of each activity according to the sentiment score, sort the matching scores, select the top k activities with the highest matching degree, and then filter out the highest ranking activities by setting the matching threshold;

[0251] S2.3, generating ranking results, and using the results as a personalized activity recommendation list for the user;

[0252] S2.4. Collect user feedback in real time, calculate user satisfaction scores based on user feedback, and adjust subsequent activity recommendation strategies;

[0253] The AR interaction subsystem 303 is used to generate an AR scene according to the personalized activity recommendation list. The user interacts by selecting the AR scene, and the AR scene is adjusted and optimized in real time according to the real-time interaction. The following steps are performed:

[0254] S3.1. Generate AR scenes based on the personalized activity recommendation list, and dynamically adjust the AR scene content based on the real-time emotion score;

[0255] S3.2. Design a multimodal interaction method based on emotion, obtain users’ emotional feedback in real time during the user experience, and optimize the interactivity of the AR scene in real time based on this feedback. Finally, automatically record the user’s emotional state, interactive behavior, and feedback at the end of each user experience.

[0256] The AR scene optimization subsystem 304 is used to optimize the recommendation planning and experience content according to the feedback data collected in real time in the AR scene, and performs the following steps:

[0257] S4.1. Construct a comprehensive feedback score function based on user feedback data combined with emotional state deviation and nonlinear adjustment factors to calculate the comprehensive feedback score;

[0258] S4.2. Adjust the priority of the next recommended activity list based on the comprehensive feedback score. If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted;

[0259] S4.3, adjusting the content complexity and interaction mode of the AR scene in real time according to the adjusted recommendation strategy, and monitoring the user's emotional state in real time for storage and analysis;

[0260] The activity optimization subsystem 305 is used to dynamically adjust the recommendation strategy by combining the user's historical preference data, the user's emotional feedback and the particularity of the travel scene, and performs the following steps:

[0261] S5.1, preprocess the user's historical preference data, comprehensive feedback score and current experience emotional state score, add exponential decay terms based on the preprocessed data to build a personalized priority adjustment model, and dynamically adjust the recommendation priority of each activity; wherein, the exponential decay term is used to measure the difference between the current user's emotional state and the emotional matching degree of the activity, and the priority is reduced as the difference increases.

[0262] S5.2. Sort the candidate activities according to the results output by the personalized priority adjustment model and select the top k best activities to generate a personalized recommendation list;

[0263] S5.3. Design an emotion fluctuation adjustment mechanism to deal with the situation where the user's emotional state fluctuates drastically, and adjust the recommendation strategy when the user's emotion fluctuates greatly;

[0264] S5.4. Set up an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation;

[0265] The system optimization subsystem 306 is used to integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes a personalized recommendation priority list, user feedback scores, government resource data and merchant service data.

[0266] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0267] In addition, for technical details that are not described in detail in this embodiment, reference can be made to the parameter operation method provided in any embodiment of the present invention, and will not be repeated here.

[0268] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0269] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0270] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0271] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A tourism information publishing and management method based on artificial intelligence and big data, characterized in that: The method comprises: S1. Collect and classify the user’s emotional state monitored in real time, including: S1.

1. Use mobile applications and wearable devices to collect users’ physiological and social data in real time; S1.2, preprocess the collected data and extract sentiment features; S1.3, using the sentiment classification model based on support vector machine to calculate the sentiment score according to the extracted sentiment features; S1.4, setting a threshold according to the calculated emotion score to classify the emotion state; S2. Recommend suitable travel activities to the user based on the user's emotional state, including: S2.1, constructing a database containing multiple tourism activities; wherein the data set includes activity ID, activity description, suitable emotion type and emotion suitability score; S2.2, calculate the matching score of each activity according to the sentiment score, sort the matching scores, select the top k activities with the highest matching degree, and then filter out the highest ranking activities by setting the matching threshold; S2.3, generating ranking results, and using the results as a personalized activity recommendation list for the user; S2.

4. Collect user feedback in real time, calculate user satisfaction scores based on user feedback, and adjust subsequent activity recommendation strategies; S3. Generate an AR scene based on the personalized activity recommendation list. Users interact by selecting the AR scene, and adjust and optimize the AR scene in real time based on the real-time interaction, including: S3.

1. Generate AR scenes based on the personalized activity recommendation list, and dynamically adjust the AR scene content based on the real-time emotion score; S3.

2. Design a multimodal interaction method based on emotion, obtain users’ emotional feedback in real time during the user experience, and optimize the interactivity of the AR scene in real time based on this feedback. Finally, automatically record the user’s emotional state, interactive behavior, and feedback at the end of each user experience. S4. Optimize recommendation planning and experience content based on real-time feedback data collected in AR scenarios, including: S4.

1. Construct a comprehensive feedback score function based on user feedback data combined with emotional state deviation and nonlinear adjustment factors to calculate the comprehensive feedback score; S4.

2. Adjust the priority of the next recommended activity list based on the comprehensive feedback score. If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted; S4.3, adjusting the content complexity and interaction mode of the AR scene in real time according to the adjusted recommendation strategy, and monitoring the user's emotional state in real time for storage and analysis; S5. Dynamically adjust the recommendation strategy based on the user's historical preference data, user's emotional feedback, and the particularity of the travel scenario, including: S5.1, preprocess the user's historical preference data, comprehensive feedback score and current experience emotional state score, add exponential decay terms based on the preprocessed data to build a personalized priority adjustment model, and dynamically adjust the recommendation priority of each activity; wherein, the exponential decay term is used to measure the difference between the current user's emotional state and the emotional matching degree of the activity, and the priority is reduced as the difference increases. S5.

2. Sort the candidate activities according to the results output by the personalized priority adjustment model and select the top k best activities to generate a personalized recommendation list; S5.

3. Design an emotion fluctuation adjustment mechanism to deal with the situation where the user's emotional state fluctuates drastically, and adjust the recommendation strategy when the user's emotion fluctuates greatly; S5.

4. Set up an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation; S6. Integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes personalized recommendation priority list, user feedback score, government resource data and merchant service data.

2. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The sentiment score is calculated as follows: Among them, S represents the sentiment score, which is used to represent the user's emotional state; m represents the number of selected features; F i represents the value of the i-th feature; W i represents the weight of the i-th feature.

3. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The matching score of each activity is calculated according to the sentiment score, which is expressed as follows: MS i =EAS i ×S+λ·f ( EAS i ,S) Among them, MS i represents the matching score of the i-th activity; EAS i represents the emotional fitness score of the i-th activity; S represents the user's emotional score; λ represents the weight factor, which adjusts the impact of activity fitness on the final score; f(EAS i ,S) represents an additional adaptation function that considers the nonlinear relationship between emotion and activity. The following is a representation of the following method for calculating the user satisfaction score based on the user feedback and adjusting the subsequent activity recommendation strategy: Among them, FS represents the user feedback score; S i represents the user satisfaction score of the i-th activity; β represents the adjustment factor used to adjust the impact of recommendation feedback; g(RAL) represents the function based on the popularity of the recommended activity.

4. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: In the process of dynamically adjusting the AR scene content according to the emotion score, the emotion adaptation coefficient α is introduced to adjust the complexity of the scene content according to the user's emotion score and scene adaptability, which is expressed as follows: Among them, AR adjusted represents the adjusted AR scene, which adapts to the user's current emotional state; α represents the emotional adaptation coefficient, which controls the sensitivity of the AR scene content to emotional changes; S represents the user's current emotional score, which represents the user's emotional feedback; μ S represents the mean of the global sentiment score; σ S Indicates the standard deviation of the sentiment score, used to standardize the sentiment value; EAS i represents the emotional fitness score of the ith activity, which is used to indicate how the scene matches the current emotional state; The multimodality includes vision, audio and gesture; wherein the adjustment formula of the multimodal interaction capability is expressed as follows: R AR =γ1·I v +γ2·I a +γ3·I g +λ·H ( S) Among them, R AR Represents the overall interactive response capability of the AR system; I v Represents the interactive strength of visual input; I a Indicates the interactive intensity of the audio input; I g represents the interaction strength of gesture input; γ1, γ2, and γ3 represent the weights used to control different input methods, which are adjusted based on the feedback of the user's current emotional state; H(S) represents the emotional state function, which reflects the nonlinear regulation of emotion on interactive operations.

5. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The feedback comprehensive scoring function is expressed as follows: CFS=w1·FS+w2·IF+w3·(S real -S prev ) 2 Among them, CFS represents the comprehensive feedback score, which reflects the user's overall feedback on the current experience; FS represents the user's subjective satisfaction score; IF represents the interaction frequency, which measures the depth of user participation; S real Indicates the emotional state score of the current experience; S prev Indicates the emotional state score of the previous experience; w1, w2, w3 represent weight parameters, which are used to adjust the influence of different feedback indicators; The priority of the recommended activity list is adjusted based on the comprehensive feedback score, as shown below: Among them, P new ( AID i ) Indicates the activity AID i The optimized priority of P prev ( AID i ) Indicates the activity AID i Initial priority set; CFS thresh represents the feedback score threshold. When CFS is lower than this threshold, it indicates that the user experience is poor. β represents the adjustment factor, which is used to control the impact of feedback on the recommendation priority. H(AID i ) represents the history function, which is used to calculate the user's past experience with the activity AID i λ represents the weight coefficient, which adjusts the proportion of historical preference in the priority.

6. The tourism information publishing and management method based on artificial intelligence and big data according to claim 5 is characterized in that: If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted, as shown below: Among them, R adjust represents the recommendation rhythm adjustment coefficient, which controls the frequency and intensity of activity recommendations; α represents the emotional state adjustment coefficient, which controls the impact of emotional fluctuations on the recommendation rhythm; γ represents the feedback adjustment coefficient, which controls the impact of feedback ratings on the recommendation rhythm; T S Indicates the regulatory time constant of the emotional state and determines the sensitivity of emotional fluctuations.

7. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The content complexity and interaction mode of the AR scene are adjusted in real time according to the adjusted recommendation strategy, as shown below: Among them, AR opt Indicates the complexity of optimized AR content; AR prev represents the complexity of the original AR content; ρ represents the adjustment coefficient, which controls the degree of content simplification; L(AR) represents the redundancy loss function of the AR content, which is used to measure the interaction complexity of the current scene; If the feedback score CFS is lower than the threshold feedback score threshold CFS thresh , simplifies AR content, reduces interaction frequency, reduces redundant information, and improves user comfort.

8. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The recommended priority calculation of the personalized priority adjustment model is expressed as follows: Among them, P opt ( AID i ) Indicates the i-th activity AID i The optimized priority of H is used for sorting and recommendation; ( AID i ) Indicates the user's historical preference weight, representing the user's past preference for the activity AID i interest; CFS represents the user's comprehensive feedback score in the last experience; CFS max represents the highest feedback score of the user in history, which is used to normalize the current feedback score; α, β, γ respectively control the influence of historical preference, feedback score and emotional state on recommendation priority; S real Indicates the user’s current emotional state; S target ( AID i ) Representation and activity AID i The most suitable emotional state; The exponential decay term is exp(-|S real -S target (AID i )|), which is used to measure the difference between the current user's emotional state and the activity's emotional match, and quickly reduce the priority as the difference increases; The emotional fluctuation regulation mechanism is expressed as follows: Among them, P adjust ( AID i ) represents the adjusted activity priority, reflecting the change in priority after emotional fluctuations; θ represents the adjustment coefficient, controlling the impact of emotional fluctuations on priority; |S real -S target (AID i )| represents the difference between the user’s current emotional state and the target emotional state of the recommended activity.

9. The tourism information publishing and management method based on artificial intelligence and big data according to claim 1 is characterized in that: The S6 specifically includes: S6.

1. Set up a dynamic resource matching model based on user demand and resource supply to dynamically match inbound and outbound resources during peak tourism periods; S6.2, collect user feedback in real time, adjust resource matching according to the deviation of each feedback, and design traffic and passenger flow optimization models to intelligently dispatch scenic spots by combining real-time traffic data and passenger flow prediction information; S6.

3. Upload real-time multi-level data to the hierarchical data interaction platform, and interact with local governments, businesses and user devices in real time through APIs to ensure that all parties can obtain the latest resource status and user feedback in a timely manner and make corresponding scheduling adjustments; S6.

4. Record and analyze feedback data and resource allocation history to identify bottlenecks in long-term resource allocation and optimize resource utilization efficiency.

10. A tourism information publishing and management system based on artificial intelligence and big data, characterized in that: The system comprises: The emotion classification subsystem is used to collect and classify the user's emotional state monitored in real time, and performs the following steps: S1.

1. Use mobile applications and wearable devices to collect users’ physiological and social data in real time; S1.2, preprocess the collected data and extract sentiment features; S1.3, using the sentiment classification model based on support vector machine to calculate the sentiment score according to the extracted sentiment features; S1.4, setting a threshold according to the calculated emotion score to classify the emotion state; The personalized recommendation subsystem is used to recommend suitable travel activities to users based on their emotional state. It performs the following steps: S2.1, constructing a database containing multiple tourism activities; wherein the data set includes activity ID, activity description, suitable emotion type and emotion suitability score; S2.2, calculate the matching score of each activity according to the sentiment score, sort the matching scores, select the top k activities with the highest matching degree, and then filter out the highest ranking activities by setting the matching threshold; S2.3, generating ranking results, and using the results as a personalized activity recommendation list for the user; S2.

4. Collect user feedback in real time, calculate user satisfaction scores based on user feedback, and adjust subsequent activity recommendation strategies; The AR interaction subsystem is used to generate AR scenes based on the personalized activity recommendation list. Users interact by selecting AR scenes, and the AR scenes are adjusted and optimized in real time based on real-time interactions. The following steps are performed: S3.

1. Generate AR scenes based on the personalized activity recommendation list, and dynamically adjust the AR scene content based on the real-time emotion score; S3.

2. Design a multimodal interaction method based on emotion, obtain users’ emotional feedback in real time during the user experience, and optimize the interactivity of the AR scene in real time based on this feedback. Finally, automatically record the user’s emotional state, interactive behavior, and feedback at the end of each user experience. The AR scene optimization subsystem is used to optimize the recommendation planning and experience content based on the real-time feedback data collected in the AR scene. The following steps are performed: S4.

1. Construct a comprehensive feedback score function based on user feedback data combined with emotional state deviation and nonlinear adjustment factors to calculate the comprehensive feedback score; S4.

2. Adjust the priority of the next recommended activity list based on the comprehensive feedback score. If the user's emotional state is detected to have an opposite emotional fluctuation, the recommendation strategy will be automatically adjusted; S4.3, adjusting the content complexity and interaction mode of the AR scene in real time according to the adjusted recommendation strategy, and monitoring the user's emotional state in real time for storage and analysis; The activity optimization subsystem is used to dynamically adjust the recommendation strategy by combining the user's historical preference data, the user's emotional feedback, and the particularity of the travel scene. It performs the following steps: S5.1, preprocess the user's historical preference data, comprehensive feedback score and current experience emotional state score, add exponential decay terms based on the preprocessed data to build a personalized priority adjustment model, and dynamically adjust the recommendation priority of each activity; wherein, the exponential decay term is used to measure the difference between the current user's emotional state and the emotional matching degree of the activity, and the priority is reduced as the difference increases. S5.

2. Sort the candidate activities according to the results output by the personalized priority adjustment model and select the top k best activities to generate a personalized recommendation list; S5.

3. Design an emotion fluctuation adjustment mechanism to deal with the situation where the user's emotional state fluctuates drastically, and adjust the recommendation strategy when the user's emotion fluctuates greatly; S5.

4. Set up an adaptive learning mechanism to continuously optimize the personalized recommendation strategy based on the actual feedback of each recommendation; The system optimization subsystem is used to integrate multi-level data, real-time resource optimization and dynamic scheduling; wherein the multi-level data includes personalized recommendation priority lists, user feedback scores, government resource data and merchant service data.

Images