AI-based cultural tourism content personalized recommendation system

Abstract

The invention discloses an AI-based cultural tourism content personalized recommendation system, which relates to the technical field of tourism planning and comprises a user information acquisition module, a cultural tourism content database information storage module, a recommendation engine module, an emotion analysis module, a real-time recommendation updating module and a multi-channel display module. According to the method, basic information, tourism preference, historical behaviors, geographic positions and social media data of users are collected, personalized portraits are generated and accurately recommended, sentiment analysis, real-time positions and dynamic behaviors are combined, the recommended content is continuously optimized by the system to adapt to the changing requirements of the users, and the user experience is improved. Through multi-channel display, personalized labels and clustering analysis, the system provides customized tourism suggestions, supports collaborative filtering and content recommendation strategies, real-time feedback and a self-optimization mechanism to improve the precision, adapts to intelligent equipment, recommends tourism contents according to environmental changes, and improves the user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of tourism planning, and in particular to an AI-based personalized recommendation system for cultural tourism content. Background Art

[0002] With the popularization of the Internet and mobile devices, cultural tourism has become a lifestyle that more and more people around the world are paying attention to. When planning cultural tourism, people not only need to choose a suitable tourist destination, but also hope to find travel activities and content that suits their personal interests and needs. Traditional travel recommendation methods often have problems such as information overload and lack of personalization, making it difficult to provide accurate and personalized travel recommendations for each user. In order to improve the quality of cultural tourism services and user experience, more and more travel platforms have begun to introduce recommendation systems based on artificial intelligence, using advanced technologies such as big data analysis, machine learning algorithms, natural language processing, and sentiment analysis to provide users with personalized travel recommendations.

[0003] Although existing cultural tourism recommendation systems are able to recommend content based on user portraits, they still face the problem that user behavior and interests will change with time, location and context. Traditional recommendation systems often lack timely adaptation to such dynamic changes, and different user devices and platforms have different interaction characteristics. How to ensure that the display effect of recommended content on various devices is consistent with the user's interaction needs is a major challenge in system design. Summary of the invention

[0004] In order to solve the above technical problems, a personalized recommendation system for cultural tourism content based on AI is provided. This technical solution solves the above problems.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] An AI-based personalized recommendation system for cultural tourism content, including:

[0007] User information collection module: The user information collection module is used to collect the user's basic information, travel preferences, historical behavior data, geographic location and social media data to generate a personalized user portrait;

[0008] Cultural tourism content database information storage module: The cultural tourism content database information storage module is used to store various types of cultural tourism content and classify them by tags and categories;

[0009] Recommendation engine module: The recommendation engine module is used to make personalized recommendations based on machine learning algorithms, using user portraits and historical behavior data to recommend cultural tourism content that is most suitable for users;

[0010] Sentiment analysis module: The sentiment analysis module is used to understand the user's emotional tendencies in real time, adjust the recommendation strategy, and optimize the personalized recommendation effect by performing sentiment analysis on user feedback, comments, and social media content;

[0011] Real-time recommendation update module: The real-time recommendation update module is used to adjust the recommended content according to the user's real-time location changes and dynamic behavior during the travel process;

[0012] Multi-channel display module: Display recommended content to users through mobile applications, website platforms, voice assistants and smart device channels.

[0013] Preferably, the user information collection module specifically includes:

[0014] User basic information collection unit: collects user personal information, including basic information such as age, gender, occupation, education level, language preference and interests;

[0015] Tourism preference collection unit: collects users’ specific preferences for cultural tourism related activities, destinations and experiences;

[0016] Historical behavior data collection unit: collects and analyzes users’ interactive behaviors and historical activity data within the platform;

[0017] Geographic location data collection unit: obtains real-time GPS data from user devices, recommends nearby tourist attractions, activities and services based on the user's current location, tracks the user's travel path, destination, and movement between attractions, generates real-time recommendations related to the current location, obtains the user's movement trajectory, understands the places they have visited and potential points of interest, and stores and processes historical location data;

[0018] Social media data collection unit: obtain users' publicly shared content, interaction data and emotional tendencies through social media platforms to further improve user portraits;

[0019] User interaction behavior data collection unit: captures all interaction information between users and the platform, conducts feedback analysis on user interactions on the interface, and identifies changes in user expectations and preferences;

[0020] Data preprocessing and cleaning unit: preprocess the collected user data, remove duplicate data records, delete missing user data, and normalize the data;

[0021] Personalized portrait generation unit: Generates a personalized portrait of the user by integrating all collected user data as the input of the subsequent recommendation engine. According to the behavior patterns of different users, labels are assigned to them and the personalized portrait is dynamically updated.

[0022] Preferably, the personalized portrait generation unit specifically includes:

[0023] Represent the user's personalized portrait as a vector;

[0024] The frequency of users' visits to scenic spots and their stay time are expressed by behavior frequency and duration, and feature vectors are dynamically generated;

[0025] Based on the feature vector, the feature vectors of various behavioral data are weighted by weighted average method to obtain the final user portrait. The weighted average formula is:

[0026]

[0027] Where, X u For user portrait, f i (B u ) is based on user behavior data B u The i-th feature extracted, w i Represents the importance weight of the feature, and n is the number of feature dimensions;

[0028] Assign tags to different users according to their behavior patterns, classify users' interests through clustering algorithms, and dynamically generate tags related to users' interests;

[0029] Dynamically update personalized portraits when user behaviors and interests change.

[0030] Preferably, the step of assigning tags to users according to their different behavior patterns, classifying the interests of users by a clustering algorithm, and dynamically generating tags related to the interests of users specifically includes:

[0031] By minimizing the distance between each user behavior and its cluster center, the best cluster is found. According to the clustering results, a dynamic label is generated for each category. The cluster analysis formula is:

[0032]

[0033] In the formula, J is the sum of the squares of the distances from all sample points to their cluster centers, k is the number of clusters, and c is i is the i-th cluster, x j is the jth sample point, μ i is the center point of the i-th cluster, ‖x j -μ i ‖ 2 Represents the sample point x j To the cluster center μ i The square of the Euclidean distance.

[0034] Preferably, when the user's behavior and interests change, dynamically updating the personalized portrait specifically includes:

[0035] Based on the changes in user behavior and interests, a dynamic user portrait update model is established, where the dynamic user portrait update model expression is:

[0036] X u (t+1)=α·X u (t)+(1-α)·f u (B u (t+1))

[0037] Where, X u (t+1) represents the updated user profile of user u at time t+1, X u (t) is the old user portrait of user u at time t, α is a balance coefficient, f u B u (t+1) is the behavior set B of user u at time t+1 u (t+1) The new interest feature generated, B u (t+1) is the behavior set of user u at time t+1.

[0038] Preferably, the recommendation engine module specifically includes:

[0039] User portrait acquisition unit: obtain the latest user portrait data;

[0040] Content feature extraction and modeling unit: extract features of cultural tourism content, including cultural background, geographic location, tourism type, and user ratings, comments and label information of the content, and use natural language processing technology to process text description data;

[0041] Personalized recommendation model unit: Based on user collaborative filtering, find users with similar historical behaviors to the target user, recommend content that similar users like, and recommend other items similar to items that the user has liked before;

[0042] Content recommendation unit: recommends cultural tourism content that matches the user's interests based on the user's preferences for content features in the user's profile;

[0043] Personalized recommendation generation unit: predicts the user's potential interest score for different cultural tourism contents based on user portraits and content features, and generates a personalized recommendation list based on the predicted score;

[0044] Recommendation result evaluation and optimization unit: Real-time evaluation of recommendation effects through A / B testing, user click-through rate, and conversion rate indicators, and adjustment of models based on feedback data.

[0045] Preferably, the personalized recommendation generating unit specifically includes:

[0046] Based on user collaborative filtering, we predict the content that the user may like by finding other users with similar behaviors to the target user. The prediction formula is:

[0047]

[0048] In the formula, r ui is the predicted score of user u for item i, N(u) is the set of users with similar behaviors to user u, and w uv is the similarity between user u and user v, r ui is the rating of item i by user u;

[0049] Generate a list of recommendations based on the predicted scores.

[0050] Preferably, the recommendation result evaluation and optimization unit specifically includes:

[0051] Evaluate the performance of the recommendation system through A / B testing, CTR and CVR indicators, and adjust the model based on the evaluation results;

[0052] The model adjustment includes hyperparameter adjustment and feature optimization, and the recommendation system is continuously optimized through online learning and real-time feedback data.

[0053] Preferably, it is characterized in that the multi-channel display module specifically includes:

[0054] Content display design and adaptation unit: After the recommended content is generated, it is displayed and interacted with in different channels, and card-based display is performed based on the recommended content, and responsive design is performed according to different screen sizes;

[0055] SEO optimization unit: optimize the display of recommended content through SEO strategies to increase content exposure and user discoverability;

[0056] Voice assistant adaptation unit: Use voice prompts in the form of short sentences and lists to guide users to perform subsequent operations, interact with users through voice, and support users to ask questions and directly execute recommended behaviors;

[0057] Smart device adaptation unit: adapt to a variety of smart devices and display recommended content through multiple channels;

[0058] Real-time feedback and optimization unit: optimize through real-time user feedback, collect user interaction behaviors, and provide real-time feedback on recommendation effects;

[0059] Personalized recommendation strategy scheduling unit: Make goal-oriented personalized recommendations based on the characteristics of each channel and the user's interaction habits, and coordinate and optimize the recommended content on other channels based on the user's behavior on one channel.

[0060] Preferably, the smart device adaptation unit specifically includes:

[0061] Multimedia display: In smart devices, recommended content is displayed in the form of video, pictures and sound;

[0062] Sensing and environmental interaction: The smart home system recommends appropriate device controls based on the room’s temperature and light data, including adjusting air conditioning and lighting.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention proposes to generate personalized portraits and make accurate recommendations by collecting users' basic information, travel preferences, historical behaviors, geographic locations and social media data. Combined with sentiment analysis, real-time location and dynamic behaviors, the system continuously optimizes recommended content to adapt to users' changing needs. Through multi-channel display, personalized tags and cluster analysis, the system provides customized travel suggestions, supports collaborative filtering and content recommendation strategies, and improves accuracy through real-time feedback and self-optimization mechanisms. At the same time, it adapts to smart devices, recommends travel content according to environmental changes, and improves user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a system framework diagram of the present invention;

[0066] Figure 2 This is an internal system framework diagram of the user information collection module in the present invention;

[0067] Figure 3 This is an internal system framework diagram of the recommendation engine module in the present invention;

[0068] Figure 4 This is an internal system framework diagram of the multi-channel display module in the present invention. DETAILED DESCRIPTION

[0069] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0070] Reference Figure 1As shown, a personalized recommendation system for cultural tourism content based on AI includes: a user information collection module: the user information collection module is used to collect the user's basic information, travel preferences, historical behavior data, geographic location and social media data, and generate a personalized user portrait; a cultural tourism content database information storage module: the cultural tourism content database information storage module is used to store a variety of cultural tourism content, and classify them by tags and categories; a recommendation engine module: the recommendation engine module is used to make personalized recommendations based on machine learning algorithms, using user portraits and historical behavior data, and recommend the most suitable cultural tourism content for users; a sentiment analysis module: the sentiment analysis module is used to understand the user's emotional tendencies in real time, adjust the recommendation strategy, and optimize the personalized recommendation effect by performing sentiment analysis on user feedback, comments, and social media content; a real-time recommendation update module: the real-time recommendation update module is used to adjust the recommended content according to the user's real-time location changes and dynamic behavior during the travel process; a multi-channel display module: the recommended content is displayed to the user through mobile applications, website platforms, voice assistants, and smart device channels.

[0071] Reference Figure 2 As shown, the user information collection module specifically includes:

[0072] User basic information collection unit: collects user personal information, including basic information such as age, gender, occupation, education level, language preference and interests;

[0073] Tourism preference collection unit: collects users’ specific preferences for cultural tourism related activities, destinations and experiences;

[0074] Historical behavior data collection unit: collects and analyzes users’ interactive behaviors and historical activity data within the platform;

[0075] Geographic location data collection unit: obtains real-time GPS data from user devices, recommends nearby tourist attractions, activities and services based on the user's current location, tracks the user's travel path, destination, and movement between attractions, generates real-time recommendations related to the current location, obtains the user's movement trajectory, understands the places they have visited and potential points of interest, and stores and processes historical location data;

[0076] Social media data collection unit: obtain users' publicly shared content, interaction data and emotional tendencies through social media platforms to further improve user portraits;

[0077] User interaction behavior data collection unit: captures all interaction information between users and the platform, conducts feedback analysis on user interactions on the interface, and identifies changes in user expectations and preferences;

[0078] Data preprocessing and cleaning unit: preprocess the collected user data, remove duplicate data records, delete missing user data, and normalize the data;

[0079] Personalized portrait generation unit: Generates a personalized portrait of the user by integrating all collected user data, which is used as the input of the subsequent recommendation engine. It assigns labels to different users according to their behavior patterns and dynamically updates the personalized portrait.

[0080] First, duplicate records are removed to eliminate redundant data and ensure data consistency. Then, missing user information is processed by filling or deleting strategies. Finally, the data is normalized to ensure that data of different scales can be compared within the same range.

[0081] The personalized portrait generation unit specifically includes:

[0082] Represent the user's personalized portrait as a vector;

[0083] The frequency of users' visits to scenic spots and their stay time are expressed by behavior frequency and duration, and feature vectors are dynamically generated;

[0084] Based on the feature vector, the feature vectors of various behavioral data are weighted by weighted average method to obtain the final user portrait. The weighted average formula is:

[0085]

[0086] Where, X u For user portrait, f i (B u ) is based on user behavior data B u The i-th feature extracted, w i Represents the importance weight of the feature, and n is the number of feature dimensions;

[0087] Assign tags to different users according to their behavior patterns, classify users' interests through clustering algorithms, and dynamically generate tags related to users' interests;

[0088] Dynamically update personalized portraits when user behaviors and interests change.

[0089] The steps of assigning tags to users according to their different behavior patterns, classifying users' interests by clustering algorithms, and dynamically generating tags related to users' interests specifically include:

[0090] By minimizing the distance between each user behavior and its cluster center, the best cluster is found. According to the clustering results, a dynamic label is generated for each category. The cluster analysis formula is:

[0091]

[0092] In the formula, J is the sum of the squares of the distances from all sample points to their cluster centers, k is the number of clusters, and c is i is the i-th cluster, x j is the jth sample point, μ i is the center point of the i-th cluster, ‖x j -μ i ‖ 2 Represents the sample point x j To the cluster center μ i The square of the Euclidean distance.

[0093] When the user's behavior and interests change, the dynamic update of the personalized portrait specifically includes:

[0094] Based on the changes in user behavior and interests, a dynamic user portrait update model is established, where the dynamic user portrait update model expression is:

[0095] X u (t+1)=α·X u (t)+(1-α)·f u (B u (t+1))

[0096] In the formula, X u (t+1) represents the updated user profile of user u at time t+1, X u (t) is the old user portrait of user u at time t, α is a balance coefficient, f u B u (t+1) is the behavior set B of user u at time t+1 u (t+1) The new interest feature generated, B u (t+1) is the behavior set of user u at time t+1.

[0097] Reference Figure 3 As shown, the recommendation engine module specifically includes:

[0098] User portrait acquisition unit: obtain the latest user portrait data;

[0099] Content feature extraction and modeling unit: extract features of cultural tourism content, including cultural background, geographic location, tourism type, and user ratings, comments and label information of the content, and use natural language processing technology to process text description data;

[0100] Personalized recommendation model unit: Based on user collaborative filtering, find users with similar historical behaviors to the target user, recommend content that similar users like, and recommend other items similar to items that the user has liked before;

[0101] Content recommendation unit: recommends cultural tourism content that matches the user's interests based on the user's preferences for content features in the user's profile;

[0102] Personalized recommendation generation unit: predicts the user's potential interest score for different cultural tourism contents based on user portraits and content features, and generates a personalized recommendation list based on the predicted score;

[0103] Recommendation result evaluation and optimization unit: Real-time evaluation of recommendation effects through A / B testing, user click-through rate, and conversion rate indicators, and adjustment of models based on feedback data.

[0104] The personalized recommendation generating unit specifically includes:

[0105] Based on user collaborative filtering, we predict the content that the user may like by finding other users with similar behaviors to the target user. The prediction formula is:

[0106]

[0107] In the formula, is the predicted score of user u for item i, N(u) is the set of users with similar behaviors to user u, and w uv is the similarity between user u and user v, r ui is the rating of item i by user u;

[0108] Generate a list of recommendations based on the predicted scores.

[0109] The recommendation result evaluation and optimization unit specifically includes:

[0110] Evaluate the performance of the recommendation system through A / B testing, CTR and CVR indicators, and adjust the model based on the evaluation results;

[0111] The model adjustment includes hyperparameter adjustment and feature optimization, and the recommendation system is continuously optimized through online learning and real-time feedback data;

[0112] The model adjustment includes hyperparameter adjustment and feature optimization. Hyperparameter adjustment refers to improving the performance and generalization ability of the model by adjusting the parameters in the model, including the learning rate, regularization coefficient, and tree depth. Feature optimization involves analyzing and selecting the most relevant features and removing redundant features to ensure that the model can more accurately capture the key information in the data. The online learning mechanism allows the system to dynamically adjust the model as user behavior changes and new data flows in, avoiding static models from becoming outdated or invalid. Real-time feedback data includes users' click-through rates, conversion rates, ratings, and comments, helping the system to promptly identify which recommendation strategies are effective and which need adjustment.

[0113] Reference Figure 4 As shown, the multi-channel display module specifically includes:

[0114] Content display design and adaptation unit: After the recommended content is generated, it is displayed and interacted with in different channels, and card-based display is performed based on the recommended content, and responsive design is performed according to different screen sizes;

[0115] SEO optimization unit: optimize the display of recommended content through SEO strategies to increase content exposure and user discoverability;

[0116] Voice assistant adaptation unit: Use voice prompts in the form of short sentences and lists to guide users to perform subsequent operations, interact with users through voice, and support users to ask questions and directly execute recommended behaviors;

[0117] Smart device adaptation unit: adapt to a variety of smart devices and display recommended content through multiple channels;

[0118] Real-time feedback and optimization unit: optimize through real-time user feedback, collect user interaction behaviors, and provide real-time feedback on recommendation effects;

[0119] Personalized recommendation strategy scheduling unit: Make goal-oriented personalized recommendations based on the characteristics of each channel and the user's interaction habits, and coordinate and optimize the recommended content on other channels based on the user's behavior on one channel.

[0120] The smart device adaptation unit specifically includes:

[0121] Multimedia display: In smart devices, recommended content is displayed in the form of video, pictures and sound;

[0122] Sensing and environmental interaction: The smart home system recommends appropriate device controls based on the room’s temperature and light data, including adjusting air conditioning and lighting.

[0123] In summary, the advantages of the present invention are:

[0124] The system comprehensively collects users' basic information, travel preferences, historical behavior data, geographic location and social media data to build a personalized profile of the user, and generates accurate cultural tourism recommendations based on the profile. Personalized recommendations can not only provide travel content that suits the user's interests based on the user's basic information, but also make dynamic adjustments based on factors such as the user's behavior changes, emotional tendencies, and real-time geographic location in different scenarios, thereby improving the relevance and accuracy of the recommendations.

[0125] By analyzing user feedback, comments, and social media content, the system can capture user emotional changes in real time, optimize recommendation strategies, and make recommended content more in line with users' current psychological needs. This emotional perception mechanism helps improve user experience and reduce the indifference and irrelevance of the recommendation system.

[0126] The system combines the real-time geographic location and dynamic behavior during the travel process to update the user's recommended content in a timely manner. Whether the user is traveling or near a scenic spot, the system can generate more accurate recommended content based on the user's real-time location and behavior data to ensure that the user receives timely and effective travel suggestions;

[0127] This system displays recommended content to users through multiple display channels and optimizes each channel. The system supports voice interaction, card-style display, multimedia content presentation, etc. to ensure that different users have a consistent and personalized experience on different devices.

[0128] By analyzing user behavior data through clustering algorithms, the system can classify user interests and generate relevant tags for them based on dynamically changing interests. This tag management enables the system to capture changes in user interests in a timely manner and provide users with more accurate recommended content.

[0129] The system provides various types of cultural tourism content through multi-level recommendation strategies based on user collaborative filtering, content feature recommendation, and personalized recommendation models, ensuring that each user can receive tailored recommendations in all aspects;

[0130] The recommendation system optimizes itself through real-time user feedback, and continuously adjusts and improves the recommendation algorithm through online learning and feature optimization to improve the accuracy and effectiveness of recommendations;

[0131] The system not only supports various traditional devices, but can also adapt to smart devices, and recommend corresponding travel content or activities to users based on the device's environmental data, thereby improving the overall travel experience.

[0132] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A personalized recommendation system for cultural tourism content based on AI, characterized in that: include: User information collection module: The user information collection module is used to collect the user's basic information, travel preferences, historical behavior data, geographic location and social media data to generate a personalized user portrait; Cultural tourism content database information storage module: The cultural tourism content database information storage module is used to store various types of cultural tourism content and classify them by tags and categories; Recommendation engine module: The recommendation engine module is used to make personalized recommendations based on machine learning algorithms, using user portraits and historical behavior data to recommend cultural tourism content that is most suitable for users; Sentiment analysis module: The sentiment analysis module is used to understand the user's emotional tendencies in real time, adjust the recommendation strategy, and optimize the personalized recommendation effect by performing sentiment analysis on user feedback, comments, and social media content; Real-time recommendation update module: The real-time recommendation update module is used to adjust the recommended content according to the user's real-time location changes and dynamic behavior during the travel process; Multi-channel display module: Display recommended content to users through mobile applications, website platforms, voice assistants and smart device channels.

2. According to claim 1, a personalized recommendation system for cultural tourism content based on AI is characterized in that: The user information collection module specifically includes: User basic information collection unit: collects user personal information, including basic information such as age, gender, occupation, education level, language preference and interests; Tourism preference collection unit: collects users’ specific preferences for cultural tourism related activities, destinations and experiences; Historical behavior data collection unit: collects and analyzes users’ interactive behaviors and historical activity data within the platform; Geographic location data collection unit: obtains real-time GPS data from user devices, recommends nearby tourist attractions, activities and services based on the user's current location, tracks the user's travel path, destination, and movement between attractions, generates real-time recommendations related to the current location, obtains the user's movement trajectory, understands the places they have visited and potential points of interest, and stores and processes historical location data; Social media data collection unit: obtain users' publicly shared content, interaction data and emotional tendencies through social media platforms to further improve user portraits; User interaction behavior data collection unit: captures all interaction information between users and the platform, conducts feedback analysis on user interactions on the interface, and identifies changes in user expectations and preferences; Data preprocessing and cleaning unit: preprocess the collected user data, remove duplicate data records, delete missing user data, and normalize the data; Personalized portrait generation unit: Generates a personalized portrait of the user by integrating all collected user data as the input of the subsequent recommendation engine. According to the behavior patterns of different users, labels are assigned to them and the personalized portrait is dynamically updated.

3. According to claim 2, a personalized recommendation system for cultural tourism content based on AI is characterized in that: The personalized portrait generation unit specifically includes: Represent the user's personalized portrait as a vector; The frequency of users' visits to scenic spots and their stay time are expressed by behavior frequency and duration, and feature vectors are dynamically generated; Based on the feature vector, the feature vectors of various behavioral data are weighted by weighted average method to obtain the final user portrait. The weighted average formula is: Where, X u For user portrait, f i (B u ) is based on user behavior data B u The i-th feature extracted, w i Represents the importance weight of the feature, and n is the number of feature dimensions; Assign tags to different users according to their behavior patterns, classify users' interests through clustering algorithms, and dynamically generate tags related to users' interests; Dynamically update personalized portraits when user behaviors and interests change.

4. According to claim 3, the AI-based personalized cultural tourism content recommendation system is characterized in that: The steps of assigning tags to users according to their different behavior patterns, classifying users' interests by clustering algorithms, and dynamically generating tags related to users' interests specifically include: By minimizing the distance between each user behavior and its cluster center, the best cluster is found. According to the clustering results, a dynamic label is generated for each category. The cluster analysis formula is: In the formula, J is the sum of the squares of the distances from all sample points to their cluster centers, k is the number of clusters, and c is i is the i-th cluster, x j is the jth sample point, μ i is the center point of the i-th cluster, ‖x j -μ i ‖ 2 Represents the sample point x j To the cluster center μ i The square of the Euclidean distance.

5. According to claim 4, the AI-based personalized recommendation system for cultural tourism content is characterized in that: When the user's behavior and interests change, the dynamic update of the personalized portrait specifically includes: Based on the changes in user behavior and interests, a dynamic user portrait update model is established, where the dynamic user portrait update model expression is: X u (t+1)=α·X u (t)+(1-α)·f u (B u (t+1)) Where, X u (t+1) represents the updated user profile of user u at time t+1, X u (t) is the old user portrait of user u at time t, α is a balance coefficient, f u B u (t+1) is the behavior set B of user u at time t+1 u (t+1) The new interest feature generated, B u (t+1) is the behavior set of user u at time t+1.

6. According to claim 5, the AI-based personalized recommendation system for cultural tourism content is characterized in that: The recommendation engine module specifically includes: User portrait acquisition unit: obtain the latest user portrait data; Content feature extraction and modeling unit: extract features of cultural tourism content, including cultural background, geographic location, tourism type, and user ratings, comments and label information of the content, and use natural language processing technology to process text description data; Personalized recommendation model unit: Based on user collaborative filtering, find users with similar historical behaviors to the target user, recommend content that similar users like, and recommend other items similar to items that the user has liked before; Content recommendation unit: recommends cultural tourism content that matches the user's interests based on the user's preferences for content features in the user's profile; Personalized recommendation generation unit: predicts the user's potential interest score for different cultural tourism contents based on user portraits and content features, and generates a personalized recommendation list based on the predicted score; Recommendation result evaluation and optimization unit: Real-time evaluation of recommendation effects through A / B testing, user click-through rate, and conversion rate indicators, and adjustment of models based on feedback data.

7. The AI-based personalized recommendation system for cultural tourism content according to claim 6 is characterized in that: The personalized recommendation generating unit specifically includes: Based on user collaborative filtering, we predict the content that the user may like by finding other users with similar behaviors to the target user. The prediction formula is: In the formula, is the predicted score of user u for item i, N(u) is the set of users with similar behaviors to user u, and w uv is the similarity between user u and user v, r ui is the rating of item i by user u; Generate a list of recommendations based on the predicted scores.

8. The AI-based personalized recommendation system for cultural tourism content according to claim 7, characterized in that: The recommendation result evaluation and optimization unit specifically includes: Evaluate the performance of the recommendation system through A / B testing, CTR and CVR indicators, and adjust the model based on the evaluation results; The model adjustment includes hyperparameter adjustment and feature optimization, and the recommendation system is continuously optimized through online learning and real-time feedback data.

9. The AI-based personalized recommendation system for cultural tourism content according to claim 8, characterized in that: The multi-channel display module specifically includes: Content display design and adaptation unit: After the recommended content is generated, it is displayed and interacted with in different channels, and card-based display is performed based on the recommended content, and responsive design is performed according to different screen sizes; SEO optimization unit: optimize the display of recommended content through SEO strategies to increase content exposure and user discoverability; Voice assistant adaptation unit: Use voice prompts in the form of short sentences and lists to guide users to perform subsequent operations, interact with users through voice, and support users to ask questions and directly execute recommended behaviors; Smart device adaptation unit: adapt to a variety of smart devices and display recommended content through multiple channels; Real-time feedback and optimization unit: optimize through real-time user feedback, collect user interaction behaviors, and provide real-time feedback on recommendation effects; Personalized recommendation strategy scheduling unit: Make goal-oriented personalized recommendations based on the characteristics of each channel and the user's interaction habits, and coordinate and optimize the recommended content on other channels based on the user's behavior on one channel.

10. The AI-based personalized cultural tourism content recommendation system according to claim 9, characterized in that: The smart device adaptation unit specifically includes: Multimedia display: In smart devices, recommended content is displayed in the form of video, pictures and sound; Sensing and environmental interaction: The smart home system recommends appropriate device controls based on the room’s temperature and light data, including adjusting air conditioning and lighting.

Images