Scenic spot recommendation method and system based on tourist portrait construction

Abstract

The invention discloses a tourist attraction recommendation method and system based on tourist portrait construction, and relates to the technical field of tourist attraction recommendation methods, and the method comprises the steps: collecting and cleaning basic information, social media behaviors and consumption habits of a user through employing a multi-source data integration and preprocessing method, and obtaining a preliminary user data set; analyzing the preliminary user data set by adopting a feature extraction and user portrait modeling method, extracting interest tags and other key features of the user from the preliminary user data set, and establishing a user portrait model to obtain a user portrait; recording interactive behaviors such as click, like, sharing and the like of the user in the user portrait in the system using process by adopting a real-time feedback mechanism method to obtain instant feedback data of the user; analyzing the instant feedback data of the user by adopting a dynamic adjustment algorithm method, and updating the user portrait to obtain an updated user portrait; basic information of a user is acquired by adopting a multi-source data integration and preprocessing method.

Description

Technical Field

[0001] The present invention relates to the technical field of tourist attraction recommendation methods, and in particular to a tourist attraction recommendation method and system based on visitor portrait construction. Background Art

[0002] Tourist attraction recommendation technology leverages data analysis and machine learning to provide users with personalized tourist attraction recommendations. Its core approach is to collect and analyze multi-dimensional user data, build detailed user profiles, and generate personalized attraction recommendations based on this data.

[0003] In the field of tourist attraction recommendation methods, existing tourist attraction recommendation systems often rely on limited data sources, resulting in a one-sided understanding of user interests and an inability to fully capture user preferences. In addition, many recommendation systems make recommendations based on static user portraits, ignoring the changes in user interests over time. Static methods cannot respond to users' latest needs and preference changes in a timely manner, resulting in a decrease in the relevance of recommendation results. At the same time, most existing recommendation algorithms fail to fully consider individual differences, and the recommended content is too generalized, making it difficult to meet users' personalized needs. Summary of the Invention

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

[0005] Therefore, the present invention provides a tourist attraction recommendation method based on tourist portrait construction to solve the problem that existing tourist attraction recommendation systems often rely on limited data sources, resulting in a one-sided understanding of user interests and an inability to fully capture user preferences.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for recommending tourist attractions based on tourist portraits, comprising:

[0008] Multi-source data integration and preprocessing methods are used to collect and clean users' basic information, social media behaviors, and consumption habits to obtain a preliminary user data set;

[0009] Use feature extraction and user portrait modeling methods to analyze the preliminary user data set, extract user interest tags and other key features, and build a user portrait model to obtain user portraits;

[0010] A real-time feedback mechanism is used to record the interactive behaviors of users in the user portrait, such as clicks, likes, and shares, during the process of using the system, to obtain instant feedback data from users;

[0011] Adopting a dynamic adjustment algorithm to analyze the user's immediate feedback data and update the user profile to obtain an updated user profile;

[0012] Use deep learning model training methods to conduct large-scale training on updated user profiles and historical user data, optimize model parameters, and obtain a personalized recommendation model;

[0013] A scene-aware recommendation method is used to combine time and weather factors, and a personalized recommendation list of tourist attractions is generated based on a personalized recommendation model to obtain the attraction recommendation results.

[0014] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, the multi-source data integration and preprocessing method is used to collect and clean users' basic information, social media behaviors, and consumption habits to obtain a preliminary user data set. The specific steps are as follows:

[0015] Use API interfaces or crawler technology to extract raw data from various data sources;

[0016] The raw data includes users' social media behaviors, consumption records on e-commerce platforms, and publicly available basic information;

[0017] The user's social media behavior includes likes, comments, and shares;

[0018] The consumption records of the e-commerce platform include purchase history and browsing history;

[0019] The disclosed basic information includes age, gender and occupation;

[0020] Store the collected raw data in a distributed database;

[0021] Calculate the relevance score based on the match between the data content and the user profile construction goal;

[0022] Calculate the data completeness score based on the missing data fields;

[0023] Calculate the freshness score based on the difference between the data generation time and the current time;

[0024] Design a comprehensive scoring function based on relevance score, data completeness score, and freshness score;

[0025] The retained data entries are deduplicated, and outliers in the data are detected and corrected. Finally, the data formats from different sources are unified to obtain a preliminary user data set.

[0026] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, the feature extraction and user portrait modeling method is used to analyze the preliminary user data set, extract the user's interest tags and other key features, and establish a user portrait model to obtain the user portrait. The specific steps are as follows:

[0027] Extract key feature categories from the preliminary user dataset;

[0028] The key feature categories include user interest tags, demographic information, and behavioral patterns;

[0029] For data containing text, natural language processing (NLP) technology is used to perform sentiment analysis and topic extraction;

[0030] Calculate the importance score TFIDF of each word using the TF-IDF algorithm, and select the top N keywords with the highest scores as interest tags;

[0031] Normalize the numerical data so that its value range falls between [0,1];

[0032] One-Hot Encoding of categorical fields;

[0033] Design a basic framework for user personas;

[0034] The basic framework of the user portrait includes basic attributes, interest tags and behavioral characteristics dimensions;

[0035] Cluster user feature vectors using the K-means algorithm to find the center point that best represents the user group;

[0036] Assign corresponding interest tags to each user based on user behavior data and text analysis results;

[0037] Combine the results of the above steps to generate a personalized profile for each user;

[0038] User portrait P u Represented as a vector containing multi-dimensional eigenvalues.

[0039] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, the real-time feedback mechanism method is used to record the interactive behaviors of users in the user portrait during the use of the system, such as clicks, likes, and shares, to obtain instant feedback data from the users. The specific steps are as follows:

[0040] Set up corresponding event listeners in the system's front-end interface to capture every user interaction behavior;

[0041] The captured interactive behavior data is transmitted to the backend database through the API interface;

[0042] Design instant feedback scores to quantify users’ timely feedback;

[0043] Set weights for different interaction types based on business needs;

[0044] For each interaction record, the corresponding instant feedback score F is calculated according to the instant feedback scoring formula.

[0045] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, wherein: the dynamic adjustment algorithm method is used to analyze the user's instant feedback data and update the user portrait to obtain the updated user portrait. The specific steps are:

[0046] Summarize all users’ instant feedback scores and group them by user ID to form an instant feedback dataset for each user;

[0047] Adjust the weight of interest tags in user portraits based on instant feedback scores;

[0048] Let the original interest tag weight vector be P u =[a1,a2,...,a n ], where a j Indicates the user's interest intensity in the jth interest tag.

[0049] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, the deep learning model training method is used to perform large-scale training on the updated user portrait and historical user data, optimize the model parameters, and obtain a personalized recommendation model. The specific steps are as follows:

[0050] Choose a neural network model based on the Transformer architecture;

[0051] Use the Transformer model with self-attention mechanism S-AM to capture the complex relationships between user behaviors;

[0052] The updated user profile P′ u Integrate with historical user data to form a complete training data set;

[0053] Use batch gradient descent algorithm to train the model;

[0054] At each iteration, a batch of samples is randomly selected from the training set, the loss value of the current batch is calculated, and the model parameters are adjusted according to the backpropagation algorithm;

[0055] Use the validation set to evaluate the performance of the final trained model and obtain a personalized recommendation model.

[0056] As a preferred solution of the tourist attraction recommendation method based on tourist portrait construction described in the present invention, the scene-aware recommendation method is combined with time and weather factors, and a personalized tourist attraction recommendation list is generated based on a personalized recommendation model to obtain the attraction recommendation results. The specific steps are as follows:

[0057] Get the current time information from the system or external API, and use the weather API to get real-time weather data for the user's location;

[0058] According to the influence of time and weather on scenic spot recommendations, a situational influence factor SS is designed to adjust the output results of the personalized recommendation model;

[0059] Load the trained personalized recommendation model and update the updated user portrait vector P′ u Input historical user data into the personalized recommendation model to obtain preliminary attraction recommendation scores;

[0060] Apply the situational impact factor S to the preliminary attraction recommendation score R i , adjust the final recommendation score R′ i ;

[0061] According to the adjusted recommendation score R′ i Arrange all candidate attractions in descending order to obtain a sorted attraction list;

[0062] The first N attractions are selected from the sorted list of attractions as the final recommendation results.

[0063] In a second aspect, the present invention provides a tourist attraction recommendation system based on tourist portraits, comprising:

[0064] Multi-source data processing module, user portrait modeling module, real-time feedback recording module, dynamic adjustment algorithm module, deep learning model training module, scene perception recommendation module and evaluation module;

[0065] The multi-source data processing module is used to extract raw data such as users' social media behaviors, e-commerce platform consumption records, and public basic information from various data sources using API interfaces or crawler technology, and clean, deduplicate, and standardize the data to obtain a preliminary user data set;

[0066] The user portrait modeling module is used to analyze the preliminary user data set, extract user interest tags and other key features, and build a user portrait model;

[0067] The real-time feedback recording module is used to set up event listeners in the front-end interface of the system to capture users' interactive behaviors such as clicks, likes, and shares, and transmit the interactive behavior data to the back-end database through the API interface, calculate the instant feedback score and record the user's instant feedback data;

[0068] The dynamic adjustment algorithm module is used to analyze the user's instant feedback data and adjust the weight of the interest tags in the user portrait according to the instant feedback score, thereby updating the user portrait to ensure that the user portrait can reflect the user's latest interest changes;

[0069] The deep learning model training module is used to perform large-scale training on the updated user profiles and historical user data to optimize model parameters;

[0070] The scenario-aware recommendation module is used to generate a personalized list of recommended tourist attractions based on a personalized recommendation model, taking into account time and weather factors;

[0071] The evaluation and feedback loop module is used to evaluate the effectiveness of the generated scenic spot recommendation results and collect user feedback data.

[0072] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for recommending tourist attractions based on tourist portraits as described in the first aspect of the present invention is implemented.

[0073] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for recommending tourist attractions based on tourist portraits as described in the first aspect of the present invention.

[0074] The beneficial effects of the present invention are as follows: by adopting a multi-source data integration and preprocessing method to collect and clean users' basic information, social media behaviors and consumption habits, the construction of a preliminary user data set is realized, which not only ensures that comprehensive data is obtained from multiple data sources, but also screens out high-quality data entries through the design of relevance scores, completeness scores and freshness scores. The preliminary user data set is analyzed through feature extraction and user portrait modeling methods, and users' interest tags and other key features are extracted therefrom, and a user portrait model is established to realize the construction of user portraits. Natural language processing technology is used to extract interest tags from text data, and combined with the normalization processing of numerical data and the unique hot encoding of categorical fields, a comprehensive and detailed user portrait is generated. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0076] Figure 1 This is a flow chart of the tourist attraction recommendation method based on tourist portraits in Example 1.

[0077] Figure 2 This is a schematic diagram of the tourist attraction recommendation system constructed based on tourist portraits in Example 1. DETAILED DESCRIPTION

[0078] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0079] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0080] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0081] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for recommending tourist attractions based on tourist portraits, comprising the following steps:

[0082] S1. Use multi-source data integration and preprocessing methods to collect and clean users' basic information, social media behaviors, and consumption habits to obtain a preliminary user dataset;

[0083] Going a step further, use API interfaces or crawler technology to extract raw data from various data sources;

[0084] The raw data includes users’ social media behaviors, consumption records on e-commerce platforms, and public basic information;

[0085] Users’ social media behaviors include likes, comments, and shares;

[0086] Consumption records on e-commerce platforms include purchase history and browsing history;

[0087] The basic information disclosed includes age, gender and occupation;

[0088] Store the collected raw data in a distributed database;

[0089] Calculate the relevance score based on the match between the data content and the user profile construction goal;

[0090] Calculate the data completeness score based on the missing data fields;

[0091] Calculate the freshness score based on the difference between the data generation time and the current time;

[0092] A comprehensive scoring function is designed based on the relevance score, data integrity score, and freshness score. The expression is:

[0093]

[0094] Among them, C is the comprehensive score of data quality. The higher the value, the better the data quality and the more suitable it is for inclusion in the final dataset. R is the relevance score of the data. I is the integrity score of the data. k is an adjustment factor used to control the impact of freshness. F is the freshness score of the data. t is the threshold parameter of freshness.

[0095] De-duplicate the retained data entries, detect and correct outliers in the data, and finally unify the data formats from different sources to obtain a preliminary user data set;

[0096] It should be noted that this module not only ensures the comprehensiveness and accuracy of the data, but also effectively screens out high-quality data entries through the design of a comprehensive scoring function, thereby providing a solid foundation for subsequent user portrait modeling. The comprehensive evaluation method based on relevance, completeness and freshness can dynamically adapt to the characteristics of different data sources, thereby improving the reliability and personalization of the entire recommendation system.

[0097] S2. Analyze the preliminary user data set using feature extraction and user portrait modeling methods, extract user interest tags and other key features, and build a user portrait model to obtain user portraits;

[0098] Furthermore, key feature categories are extracted from the preliminary user dataset;

[0099] Key feature categories include users’ interest tags, demographic information, and behavioral patterns;

[0100] For data containing text, natural language processing (NLP) technology is used to perform sentiment analysis and topic extraction;

[0101] The importance score TFIDF of each word is calculated using the TF-IDF algorithm, and the top N keywords with the highest scores are selected as interest tags. The expression is:

[0102]

[0103] Where w is a word, d is a document or post, TF(w,d) is the frequency of word w in document d, N is the total number of documents, and DF(w) is the number of documents containing word w.

[0104] Normalize the numerical data so that its value range falls between [0,1];

[0105] One-Hot Encoding of categorical fields;

[0106] Design a basic framework for user personas;

[0107] The basic framework of user portraits includes basic attributes, interest tags, and behavioral characteristics;

[0108] The user feature vectors are clustered using the K-means algorithm to find the center point that best represents the user group. The expression is:

[0109]

[0110] Among them, C i is the center point of the i-th cluster, S i is the set of all user feature vectors belonging to the i-th cluster, x j is the feature vector of user j;

[0111] Assign corresponding interest tags to each user based on user behavior data and text analysis results;

[0112] Combine the results of the above steps to generate a personalized profile for each user;

[0113] User portrait P u Represented as a vector containing multi-dimensional eigenvalues, the expression is:

[0114] P u =[a1,a2,...,a n ];

[0115] Among them, P u is the user portrait vector of user u, a1,a2,...,a n Represent the user's feature values in different dimensions respectively;

[0116] It should be noted that by using natural language processing technology to extract interest tags from text data, and combining the normalization of numerical data and one-hot encoding of categorical fields, this module can generate a comprehensive and detailed user portrait. The application of the K-means clustering algorithm further enhances the accuracy and interpretability of user group division, so that each user's interest preferences can be accurately captured and represented.

[0117] S3. Use a real-time feedback mechanism to record the interactive behaviors of users in the user portrait, such as clicks, likes, and shares, during the process of using the system, and obtain instant feedback data from users;

[0118] Furthermore, corresponding event listeners are set up in the front-end interface of the system to capture every interactive behavior of the user;

[0119] The captured interactive behavior data is transmitted to the backend database through the API interface;

[0120] Design an instant feedback score to quantify the user's timely feedback. The expression is:

[0121]

[0122] Among them, F is the user's immediate feedback score, W i is the weight of the i-th interaction type, T c is the current timestamp, T i is the timestamp of the i-th interaction, k is the freshness decay coefficient;

[0123] Set weights for different interaction types based on business needs;

[0124] For each interaction record, calculate the corresponding instant feedback score F according to the instant feedback scoring formula;

[0125] It should be noted that by setting up event listeners on the front-end interface and calculating instant feedback scores, this module can capture users' interactive behaviors in real time and update users' interest changes in a timely manner. This method not only improves the response speed of the recommendation system, but also better reflects users' latest needs, thereby improving the relevance of recommendation results and user experience.

[0126] S4. Analyze the user's instant feedback data using a dynamic adjustment algorithm and update the user profile to obtain an updated user profile;

[0127] Furthermore, all users’ instant feedback scores are aggregated and grouped by user ID to form an instant feedback dataset for each user;

[0128] Adjust the weight of interest tags in user portraits based on instant feedback scores;

[0129] Let the original interest tag weight vector be P u =[a1,a2,...,a n ], where a j Indicates the user's interest intensity in the jth interest tag;

[0130] Updated interest tag weight P′ u Expressed as:

[0131] P′ u [j]=P u [j]+α×F j ;

[0132] Among them, P′ u [j] is the updated weight of the j-th interest tag, α is the adjustment coefficient, and F j is the instant feedback score related to the jth interest tag;

[0133] It should be noted that this module achieves continuous tracking and optimization of user interest preferences through analysis of instant feedback data and dynamic updating of user portraits. By adjusting the weights of interest tags, it ensures that user portraits are always up to date, thereby improving the accuracy and effectiveness of personalized recommendations. The dynamic adjustment mechanism is particularly important for responding to rapidly changing user interests.

[0134] S5. Use deep learning model training methods to conduct large-scale training on the updated user profiles and historical user data, optimize model parameters, and obtain a personalized recommendation model;

[0135] Going further, choose a neural network model based on the Transformer architecture;

[0136] Use the Transformer model with self-attention mechanism S-AM to capture the complex relationships between user behaviors;

[0137] '

[0138] Integrate the updated user profile Pu with historical user data to form a complete training dataset;

[0139] Use batch gradient descent algorithm to train the model;

[0140] At each iteration, a batch of samples is randomly selected from the training set, the loss value of the current batch is calculated, and the model parameters are adjusted according to the back propagation algorithm. The expression is:

[0141]

[0142] Among them, w jis the jth model parameter, η is the learning rate, is the loss function L for parameter w j The partial derivative of

[0143] Use the validation set to evaluate the performance of the final trained model and obtain a personalized recommendation model;

[0144] It should be noted that choosing a neural network model based on the Transformer architecture and using the batch gradient descent algorithm for training can not only effectively capture the complex relationship between user behaviors, but also continuously optimize the model parameters through the backpropagation algorithm to improve prediction accuracy. The use of the validation set ensures that the model is not overfitting. The resulting personalized recommendation model can provide efficient and personalized recommendation services in practical applications.

[0145] S6. Using a scene-aware recommendation method combined with time and weather factors, and based on a personalized recommendation model, a personalized list of recommended tourist attractions is generated to obtain a recommendation result;

[0146] Going a step further, get the current time information from the system or external API, and use the weather API to get real-time weather data for the user's location;

[0147] According to the influence of time and weather on scenic spot recommendations, a scenario influence factor SS is designed to adjust the output results of the personalized recommendation model. The expression is:

[0148]

[0149] Among them, S is the scenario influence factor, T is the time influence factor, W is the weather influence factor, and γ is the adjustment factor;

[0150] Load the trained personalized recommendation model and update the updated user portrait vector P′ u Input historical user data into the personalized recommendation model to obtain preliminary attraction recommendation scores;

[0151] Apply the situational impact factor S to the preliminary attraction recommendation score R i , adjust the final recommendation score R′ i , the expression is:

[0152] R′ i =R i ×S;

[0153] Among them, R′ i is the recommended score of scenic spots after adjusting the situational impact factor, R i It is the original attraction recommendation score;

[0154] According to the adjusted recommendation score R′i Arrange all candidate attractions in descending order to obtain a sorted attraction list;

[0155] Select the top N attractions from the sorted list as the final recommendation results;

[0156] It should be noted that by designing scenario influencing factors in combination with time and weather factors, this module can flexibly adjust the score of scenic spot recommendation according to actual conditions, thereby generating a personalized recommendation list that is more in line with the user's current situation. This method not only takes into account the user's long-term interests, but also takes into account short-term scenario changes, making the recommendation results more practical and targeted.

[0157] This embodiment also provides a tourist attraction recommendation system based on tourist portraits, including:

[0158] Multi-source data processing module, user portrait modeling module, real-time feedback recording module, dynamic adjustment algorithm module, deep learning model training module, scene perception recommendation module and evaluation module;

[0159] The multi-source data processing module is used to extract raw data such as users' social media behaviors, e-commerce platform consumption records, and public basic information from various data sources using API interfaces or crawler technology. It then cleans, deduplicates, and standardizes the data to obtain a preliminary user data set.

[0160] The user portrait modeling module is used to analyze the preliminary user data set, extract user interest tags and other key features, and build a user portrait model;

[0161] The real-time feedback recording module is used to set event listeners in the system's front-end interface to capture users' interactive behaviors such as clicks, likes, and shares, and transmit the interactive behavior data to the back-end database through the API interface, calculate the instant feedback score, and record the user's instant feedback data;

[0162] Dynamic adjustment algorithm module, used to analyze users' instant feedback data and adjust the weights of interest tags in user portraits based on the instant feedback scores, thereby updating the user portraits to ensure that the user portraits can reflect the user's latest interest changes;

[0163] Deep learning model training module, used to conduct large-scale training on updated user profiles and historical user data to optimize model parameters;

[0164] A scenario-aware recommendation module, which combines time and weather factors and generates a personalized list of recommended tourist attractions based on a personalized recommendation model;

[0165] The evaluation and feedback loop module is used to evaluate the effectiveness of the generated attraction recommendation results and collect user feedback data.

[0166] This embodiment also provides a computer device, which is suitable for the method of recommending tourist attractions based on tourist portraits, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the method of recommending tourist attractions based on tourist portraits proposed in the above embodiment.

[0167] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through Wi-Fi, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0168] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for recommending tourist attractions based on tourist portraits as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0169] In summary, the present invention collects and cleans users' basic information, social media behaviors, and consumption habits by adopting multi-source data integration and preprocessing methods, thereby realizing the construction of a preliminary user data set. It not only ensures the acquisition of comprehensive data from multiple data sources, but also screens out high-quality data entries through the design of relevance scores, completeness scores, and freshness scores. The preliminary user data set is analyzed through feature extraction and user portrait modeling methods, and users' interest tags and other key features are extracted from them. A user portrait model is established to realize the construction of user portraits. Natural language processing technology is used to extract interest tags from text data, and combined with the normalization processing of numerical data and the unique hot encoding of categorical fields, a comprehensive and detailed user portrait is generated.

[0170] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for recommending tourist attractions based on tourist portraits, characterized by: include: Multi-source data integration and preprocessing methods are used to collect and clean users' basic information, social media behaviors, and consumption habits to obtain a preliminary user data set; Use feature extraction and user portrait modeling methods to analyze the preliminary user data set, extract user interest tags and other key features, and build a user portrait model to obtain user portraits; A real-time feedback mechanism is used to record the interactive behaviors of users in the user portrait, such as clicks, likes, and shares, during the process of using the system, to obtain instant feedback data from users; Adopting a dynamic adjustment algorithm to analyze the user's immediate feedback data and update the user profile to obtain an updated user profile; Use deep learning model training methods to conduct large-scale training on updated user profiles and historical user data, optimize model parameters, and obtain a personalized recommendation model; A scene-aware recommendation method is used to combine time and weather factors, and a personalized recommendation list of tourist attractions is generated based on a personalized recommendation model to obtain the attraction recommendation results.

2. The method for recommending tourist attractions based on tourist portraits according to claim 1, wherein: The multi-source data integration and preprocessing method is used to collect and clean users' basic information, social media behaviors, and consumption habits to obtain a preliminary user data set. The specific steps are as follows: Use API interfaces or crawler technology to extract raw data from various data sources; The raw data includes users' social media behaviors, consumption records on e-commerce platforms, and publicly available basic information; The user's social media behavior includes likes, comments, and shares; The consumption records of the e-commerce platform include purchase history and browsing history; The disclosed basic information includes age, gender and occupation; Store the collected raw data in a distributed database; Calculate the relevance score based on the match between the data content and the user profile construction goal; Calculate the data completeness score based on the missing data fields; Calculate the freshness score based on the difference between the data generation time and the current time; Design a comprehensive scoring function based on relevance score, data completeness score, and freshness score; The retained data entries are deduplicated, and outliers in the data are detected and corrected. Finally, the data formats from different sources are unified to obtain a preliminary user data set.

3. The method for recommending tourist attractions based on tourist portraits according to claim 2, wherein: The feature extraction and user portrait modeling method is used to analyze the preliminary user data set, extract the user's interest tags and other key features, and establish a user portrait model to obtain the user portrait. The specific steps are as follows: Extract key feature categories from the preliminary user dataset; The key feature categories include user interest tags, demographic information, and behavioral patterns; For data containing text, natural language processing (NLP) technology is used to perform sentiment analysis and topic extraction; Calculate the importance score TFIDF of each word using the TF-IDF algorithm, and select the top N keywords with the highest scores as interest tags; Normalize the numerical data so that its value range falls between [0,1]; One-Hot Encoding of categorical fields; Design a basic framework for user personas; The basic framework of the user portrait includes basic attributes, interest tags and behavioral characteristics dimensions; Cluster user feature vectors using the K-means algorithm to find the center point that best represents the user group; Assign corresponding interest tags to each user based on user behavior data and text analysis results; Combine the results of the above steps to generate a personalized profile for each user; User portrait P u Represented as a vector containing multi-dimensional eigenvalues.

4. The method for recommending tourist attractions based on tourist portraits according to claim 3, wherein: The real-time feedback mechanism method is used to record the interactive behaviors of users in the user portrait, such as clicks, likes, and shares, during the process of using the system, to obtain the user's instant feedback data. The specific steps are as follows: Set up corresponding event listeners in the system's front-end interface to capture every user interaction behavior; The captured interactive behavior data is transmitted to the backend database through the API interface; Design instant feedback scores to quantify users’ timely feedback; Set weights for different interaction types based on business needs; For each interaction record, the corresponding instant feedback score F is calculated according to the instant feedback scoring formula.

5. The method for recommending tourist attractions based on tourist portraits according to claim 4, wherein: The dynamic adjustment algorithm method is used to analyze the user's instant feedback data and update the user profile to obtain the updated user profile. The specific steps are: Summarize all users’ instant feedback scores and group them by user ID to form an instant feedback dataset for each user; Adjust the weight of interest tags in user portraits based on instant feedback scores; Let the original interest tag weight vector be P u =[a1,a2,...,a n ], where a j Indicates the user's interest intensity in the jth interest tag.

6. The method for recommending tourist attractions based on tourist portraits according to claim 5, wherein: The deep learning model training method is used to perform large-scale training on the updated user profile and historical user data, optimize the model parameters, and obtain a personalized recommendation model. The specific steps are as follows: Choose a neural network model based on the Transformer architecture; Use the Transformer model with self-attention mechanism S-AM to capture the complex relationships between user behaviors; The updated user profile P′ u Integrate with historical user data to form a complete training data set; Use batch gradient descent algorithm to train the model; At each iteration, a batch of samples is randomly selected from the training set, the loss value of the current batch is calculated, and the model parameters are adjusted according to the backpropagation algorithm; Use the validation set to evaluate the performance of the final trained model and obtain a personalized recommendation model.

7. The method for recommending tourist attractions based on tourist portraits according to claim 6, wherein: The scenario-aware recommendation method is combined with time and weather factors, and a personalized recommendation model is used to generate a personalized tourist attraction recommendation list to obtain the attraction recommendation results. The specific steps are as follows: Get the current time information from the system or external API, and use the weather API to get real-time weather data for the user's location; According to the influence of time and weather on scenic spot recommendations, a situational influence factor SS is designed to adjust the output results of the personalized recommendation model; Load the trained personalized recommendation model and update the updated user portrait vector P′ u Input historical user data into the personalized recommendation model to obtain preliminary attraction recommendation scores; Apply the situational impact factor S to the preliminary attraction recommendation score R i , adjust the final recommendation score R′ i ; According to the adjusted recommendation score R′ i Arrange all candidate attractions in descending order to obtain a sorted attraction list; The first N attractions are selected from the sorted list of attractions as the final recommendation results.

8. A tourist attraction recommendation system based on tourist portraits, based on the tourist attraction recommendation method based on tourist portraits according to any one of claims 1 to 7, characterized in that: include: Multi-source data processing module, user portrait modeling module, real-time feedback recording module, dynamic adjustment algorithm module, deep learning model training module, scene perception recommendation module and evaluation module; The multi-source data processing module is used to extract raw data such as users' social media behaviors, e-commerce platform consumption records, and public basic information from various data sources using API interfaces or crawler technology, and clean, deduplicate, and standardize the data to obtain a preliminary user data set; The user portrait modeling module is used to analyze the preliminary user data set, extract user interest tags and other key features, and build a user portrait model; The real-time feedback recording module is used to set up event listeners in the front-end interface of the system to capture users' interactive behaviors such as clicks, likes, and shares, and transmit the interactive behavior data to the back-end database through the API interface, calculate the instant feedback score and record the user's instant feedback data; The dynamic adjustment algorithm module is used to analyze the user's instant feedback data and adjust the weight of the interest tags in the user portrait according to the instant feedback score, thereby updating the user portrait to ensure that the user portrait can reflect the user's latest interest changes; The deep learning model training module is used to perform large-scale training on the updated user profiles and historical user data to optimize model parameters; The scenario-aware recommendation module is used to generate a personalized list of recommended tourist attractions based on a personalized recommendation model, taking into account time and weather factors; The evaluation and feedback loop module is used to evaluate the effectiveness of the generated scenic spot recommendation results and collect user feedback data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for recommending tourist attractions based on tourist portraits according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for recommending tourist attractions based on tourist portraits according to any one of claims 1 to 7 are implemented.

Images