Realization method for dynamic construction and personalized recommendation of user portrait fused with reinforcement learning

Abstract

The invention belongs to the technical field of artificial intelligence personalized services, and particularly relates to a reinforcement learning-fused user portrait dynamic construction and personalized recommendation implementation method, which comprises the following steps of: acquiring multi-dimensional user data; using the multi-dimensional user data to construct a multi-dimensional dynamic user portrait through a preset algorithm; based on the multi-dimensional dynamic user portrait, combining a collaborative filtering algorithm and a deep learning algorithm to generate a scene recommendation strategy; and monitoring user feedback and behavior data in real time, and optimizing the scenarized recommendation strategy by using the real-time monitored user feedback and behavior data. According to the method, the dynamic user portrait is constructed, scene recommendation is generated, and the strategy is optimized in real time by relying on reinforcement learning, so that the individuation degree, recommendation accuracy and user satisfaction of the service are remarkably improved, and the limitation of a traditional service mode is effectively broken through.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence personalized service technology, and in particular relates to a method for dynamically constructing user profiles and making personalized recommendations by integrating reinforcement learning. It is applicable to customized services in multiple scenarios such as nature exploration, parent-child education, and travel. Background Technology

[0002] Existing AI service systems generally adopt a generic content recommendation and interaction model, which makes it difficult to meet the personalized needs of users in different scenarios. For example, in educational scenarios, children of different ages have significantly different learning progress; in nature exploration or tourism scenarios, users' physical strength and interest preferences vary significantly, and the existing "one-size-fits-all" service model leads to a poor user experience. In addition, existing systems lack a real-time dynamic adjustment mechanism, making it impossible to optimize service content based on users' immediate feedback (such as changes in interests or behavioral adjustments), further limiting the improvement of user experience. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a method for dynamically constructing user profiles and making personalized recommendations by integrating reinforcement learning, so as to solve the problems of insufficient personalization and lack of real-time optimization capabilities in the prior art.

[0004] According to a first aspect of the present invention, a method for dynamically constructing user profiles and providing personalized recommendations by integrating reinforcement learning is provided, the method comprising:

[0005] Collect multi-dimensional user data; the multi-dimensional user data includes: real-time user behavior data, user preference data, and related historical data;

[0006] Using the multi-dimensional user data, a multi-dimensional dynamic user profile is constructed through a preset algorithm;

[0007] Based on the aforementioned multi-dimensional dynamic user profile, a scenario-based recommendation strategy is generated by combining collaborative filtering and deep learning algorithms.

[0008] Real-time monitoring of user feedback and behavioral data, and optimization of the contextualized recommendation strategy using the real-time monitoring of user feedback and behavioral data.

[0009] Furthermore, the contextualized recommendation strategy includes:

[0010] Strategies for nature exploration, parent-child education, and travel.

[0011] Furthermore, the step of constructing a multi-dimensional dynamic user profile using the multi-dimensional user data through a preset algorithm includes:

[0012] Feature engineering is used to extract key features from preprocessed, multi-dimensional user data.

[0013] Based on the aforementioned key features, an initial user profile is constructed using the random forest algorithm, and the user profile is updated every set time period based on newly collected real-time data to obtain a multi-dimensional dynamic user profile.

[0014] Furthermore, the preprocessing of multi-dimensional user data includes:

[0015] Use a preset data cleaning tool to remove outliers and format the data.

[0016] Furthermore, the generation of contextualized recommendation content based on the multi-dimensional dynamic user profile, combined with collaborative filtering and deep learning algorithms, includes:

[0017] Based on the feature values ​​of the physiological ability dimension and preference and interest dimension in the multi-dimensional dynamic user profile, a collaborative filtering algorithm is used to match high-rated routes of similar users in the multi-dimensional dynamic user profile.

[0018] Based on the high-scoring routes, a deep learning model is used to predict the route fit.

[0019] Furthermore, the real-time monitoring of user feedback and behavioral data, and the optimization of the contextualized strategy using the real-time monitoring of user feedback and behavioral data, include:

[0020] Using the aforementioned user feedback and behavioral data, and based on the time-series analysis of the behavioral data, sudden changes in the user's interaction time on specific types of content are identified. When the dwell time decreases beyond a set threshold, it is determined that the point of interest has shifted.

[0021] Detect changes in interest points based on behavioral data. If the user's interaction time on a specific type of content decreases by more than a set threshold, it is determined that the user's interest points have shifted.

[0022] Otherwise, it is determined that the point of interest has not shifted;

[0023] Using the results of user feedback and interest point change detection, a reinforcement learning algorithm is employed to dynamically adjust the type weights of recommended content and optimize the contextualization strategy.

[0024] Furthermore, the collection of multi-dimensional user data includes:

[0025] Data is collected through terminal devices, including mobile phones, smart bracelets, and smartwatches;

[0026] The user's real-time behavioral data includes steps taken, heart rate, duration of stay at attractions, and frequency of page interactions.

[0027] The user preference data includes interest tags, historical likes, and search keywords;

[0028] The relevant historical data includes historical exploration routes, learning content completion records, and past itinerary arrangements.

[0029] Furthermore, the deep learning model includes:

[0030] In the natural exploration scenario, the LSTM model is used to predict the physical fitness of the current route based on the user's historical physical exertion curve and route adaptation record.

[0031] In parent-child education scenarios, a CNN model is used to perform image feature mapping on the features in the user profile and generate appropriate graphic and text learning content.

[0032] In tourism scenarios, the Transformer model is used to generate time-series itinerary recommendation sequences by combining user preferences with real-time weather and attraction crowding data.

[0033] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0034] The beneficial effects of this invention are that, by integrating reinforcement learning for dynamic user profile construction and personalized recommendation schemes, it specifically addresses the problems of insufficient personalization, poor scenario adaptability, and rigid recommendation strategies in existing technologies. Specifically, this is reflected in the following aspects:

[0035] 1. Significantly enhanced personalized experience

[0036] By creating multi-dimensional dynamic user profiles (covering core dimensions such as physical abilities and preferences), combined with scenario-based recommendation strategies (customized for scenarios such as nature exploration, parent-child education, and travel), this method achieves highly personalized services tailored to each individual. Compared to existing general recommendation models, this approach can accurately match users' physical abilities, interests, learning progress, and other personalized characteristics. For example, it can recommend routes suitable for physical abilities in nature exploration and push age-appropriate learning content in parent-child education, significantly enhancing the personalized experience.

[0037] 2. Enhanced real-time optimization capabilities improve the timeliness and accuracy of recommendations.

[0038] By leveraging reinforcement learning algorithms to dynamically optimize recommendation strategies and by monitoring user feedback (explicit ratings, text reviews) and behavioral data (dwell time, interaction frequency) in real time, the system adjusts recommendation weights instantly (e.g., rapidly updating content when interests shift), thus overcoming the limitation of existing "static recommendations" that cannot respond to real-time changes in user needs. Recommendation accuracy and user satisfaction are both effectively improved.

[0039] 3. Breakthroughs in scenario adaptability and service refinement

[0040] Differentiated recommendation logics are designed to address the core needs of different scenarios: nature exploration scenarios combine AR navigation with real-time knowledge explanations; parent-child education scenarios integrate interactive games and teaching point prompts; and travel scenarios link real-time weather with segmented itinerary planning. This deep scenario-based adaptation avoids the "one-size-fits-all" service model of existing technologies, significantly improving the match between recommended content and scenario requirements.

[0041] 4. Optimization of User Profile Dynamics and Data Utilization Efficiency

[0042] By updating user profiles according to a set time period using the random forest algorithm and combining it with feature engineering to extract key contextual features (such as "parent feedback frequency" in parent-child scenarios and "average daily activity duration" in tourism scenarios), the problem of "information lag" in traditional static profiles is solved. At the same time, the experience playback mechanism of reinforcement learning improves the utilization efficiency of user feedback data, effectively enhancing the timeliness and accuracy of profile updates.

[0043] In summary, this invention achieves a leap from "passive adaptation" to "active evolution" in personalized services through the synergy of dynamic profile construction, scenario-based recommendation, and real-time optimization. It significantly improves user experience, recommendation accuracy, and scenario adaptability, demonstrating outstanding technological advancement and practical value in the field of AI-driven personalized services.

[0044] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0045] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0046] Figure 1 This is a schematic diagram illustrating a method for dynamically constructing user profiles and providing personalized recommendations using reinforcement learning, according to an exemplary embodiment. Detailed Implementation

[0047] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0048] Example 1

[0049] Please see Figure 1 , Figure 1This is a schematic diagram illustrating a method for dynamically constructing user profiles and providing personalized recommendations using reinforcement learning, according to an exemplary embodiment. The method includes:

[0050] S1. Collect multi-dimensional user data; the multi-dimensional user data includes: real-time user behavior data, user preference data, and related historical data;

[0051] S2. Using the multi-dimensional user data, construct a multi-dimensional dynamic user profile through a preset algorithm;

[0052] S3. Based on the multi-dimensional dynamic user profile, combine collaborative filtering algorithm and deep learning algorithm to generate a scenario-based recommendation strategy;

[0053] S4. Monitor user feedback and behavior data in real time, and optimize the contextualized recommendation strategy using the real-time monitoring of user feedback and behavior data.

[0054] In specific implementation, as described in step S1, the aforementioned multi-dimensional user data encompasses three core types of information: First, real-time behavioral data, collected through terminal devices such as mobile phones and smart bracelets, including physiological movement data such as step count and heart rate, as well as scene interaction data such as time spent at scenic spots and frequency of page interactions; second, user preference data, including interest tags generated based on historical behavior (such as "plant observation"), explicit preference records (likes, actively selected interest options), and implicit preference characteristics (search keywords, content dwell time ratio); and third, relevant historical record data, involving historical activity records such as past exploration routes and completion status of learning content, historical feedback such as itinerary ratings and text feedback, as well as historical status information such as device usage habits and physiological data baselines. These data collectively form the basis of dynamic user profiles and contextualized recommendations, supporting the accurate realization of personalized services.

[0055] In specific implementation, as described in step S2, multi-dimensional data is collected (user behavior, preferences, historical records, etc.), and key features (such as interests, physical strength, etc.) are extracted after data preprocessing. Machine learning algorithms are used to construct user profiles containing information such as age, learning progress, and device usage habits, and these profiles are dynamically updated based on real-time data.

[0056] In practice, the following steps are included:

[0057] S2.1 Data Preprocessing: Use data cleaning tools to remove outliers from the collected data (such as invalid data generated by erroneous operations) and format the data (such as unifying the time format and physical strength data units).

[0058] S2.2 Feature Extraction: Key features are extracted through feature engineering. For example, in the parent-child education scenario, "child's age", "parent feedback frequency" and "learning content completion rate" are extracted; in the tourism scenario, "preferred attraction type" and "average daily activity duration" are extracted.

[0059] S2.3 User Profile Construction and Update: The initial user profile is constructed using the random forest algorithm. The profile is updated every 24 hours based on newly collected real-time data (such as daily behavior and feedback) to ensure the timeliness of information.

[0060] In one embodiment, such as step S3, based on user profiles and combined with collaborative filtering and deep learning algorithms, customized recommendations are provided in scenarios such as nature exploration, parent-child education, and tourism. These recommendations include route planning, learning content, and itinerary arrangements. For example, in nature exploration, suitable routes are recommended based on the user's physical condition, and in tourism scenarios, customized itineraries are recommended based on real-time data.

[0061] Nature exploration scenario: Based on the user profile's "physical fitness level" (e.g., historical average daily walking of 3 kilometers) and "interests" (e.g., preference for plant observation), the collaborative filtering algorithm matches high-rated routes of similar users, and combines deep learning models to predict route suitability, recommending routes within 3 kilometers that include plant observation points.

[0062] For parent-child education scenarios: Based on the characteristics of "child's age 6 years old" and "learning progress at the natural science enlightenment stage" in the profile, we recommend plant cognition learning content that combines pictures and text, and add interactive games (such as "plant spot the difference"), while marking teaching points (such as "photosynthesis can be explained here").

[0063] Tourism scenario: Combining user profiles of "preference for historical sites" with real-time data (such as sunny weather on the day), a deep learning model is used to recommend customized itineraries of "museum in the morning + outdoor historical sites in the afternoon", and audio introductions of the attractions are pushed at the same time.

[0064] In practice, as shown in step S4, user feedback and behavioral data are monitored in real time, and the recommendation strategy is optimized through reinforcement learning algorithms: after a user completes an activity, the next recommendation is adjusted based on the feedback; when a change in a user's interests is detected, the recommended content is updated immediately to ensure continuous service optimization.

[0065] S41. Feedback Collection: Collect user evaluations of recommended content by setting "Satisfied / Neutral / Unsatisfied" rating buttons and text feedback boxes through the system interface; at the same time, monitor user behavior data through sensors (e.g., staying at a scenic spot for more than 30 minutes is considered a point of interest).

[0066] S42. Strategy Optimization: Using reinforcement learning algorithms, user satisfaction is used as a reward signal. If a user rates the recommended route as "satisfactory", the recommendation weight of that type of route is increased. If a user's stay time at "historical attractions" is detected to decrease sharply (e.g., from 60 minutes to 10 minutes), it is determined that the point of interest has changed, and the recommendation is immediately adjusted to a new type of attraction.

[0067] In practical implementation, the LSTM model is used in the natural exploration scenario to predict the physical fitness of the current route based on the user's historical physical exertion curve and route adaptation record.

[0068] In parent-child education scenarios, a CNN model is used to perform image feature mapping on the features in the user profile and generate appropriate graphic and text learning content.

[0069] In tourism scenarios, the Transformer model is used to generate time-series itinerary recommendation sequences by combining user preferences with real-time weather and attraction crowding data.

[0070] This invention significantly improves the personalization and accuracy of services by integrating reinforcement learning to construct dynamic user profiles and achieve personalized recommendations. The dynamic profiles built from multi-dimensional data accurately capture users' physiological capabilities and preferences, combined with scenario-based recommendation strategies (nature exploration, parent-child education, travel, etc.), avoiding the limitations of the traditional "one-size-fits-all" approach and making services more tailored to the core needs of different scenarios. Simultaneously, relying on reinforcement learning for real-time monitoring and strategy optimization of user feedback and behavioral data solves the problem of static recommendations failing to respond to real-time changes in user needs, continuously improving the timeliness and adaptability of recommendations, ultimately achieving a significant optimization of the user experience and demonstrating outstanding technological value in the field of AI-driven personalized services.

[0071] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0072] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this invention, unless otherwise stated, "a plurality of" means at least two.

[0073] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0074] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0075] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0076] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0077] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0078] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0079] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for dynamically constructing user profiles and providing personalized recommendations by integrating reinforcement learning, characterized in that... The method includes: Collect multi-dimensional user data; the multi-dimensional user data includes: real-time user behavior data, user preference data, and related historical data; Using the multi-dimensional user data, a multi-dimensional dynamic user profile is constructed through a preset algorithm; Based on the aforementioned multi-dimensional dynamic user profile, a scenario-based recommendation strategy is generated by combining collaborative filtering and deep learning algorithms. Real-time monitoring of user feedback and behavioral data, and optimization of the contextualized recommendation strategy using the real-time monitoring of user feedback and behavioral data.

2. The method according to claim 1, characterized in that, The contextualized recommendation strategy includes: Strategies for nature exploration, parent-child education, and travel.

3. The method according to claim 1, characterized in that, The step of constructing a multi-dimensional dynamic user profile using the multi-dimensional user data and a preset algorithm includes: Feature engineering is used to extract key features from preprocessed, multi-dimensional user data. Based on the aforementioned key features, an initial user profile is constructed using the random forest algorithm, and the user profile is updated every set time period based on newly collected real-time data to obtain a multi-dimensional dynamic user profile.

4. The method according to claim 3, characterized in that, Preprocessing of multi-dimensional user data, including: Use a preset data cleaning tool to remove outliers and format the data.

5. The method according to claim 1, characterized in that, The process of generating contextualized recommendation content based on the multi-dimensional dynamic user profile, combined with collaborative filtering and deep learning algorithms, includes: Based on the feature values ​​of the physiological ability dimension and preference and interest dimension in the multi-dimensional dynamic user profile, a collaborative filtering algorithm is used to match high-rated routes of similar users in the multi-dimensional dynamic user profile. Based on the high-scoring routes, a deep learning model is used to predict the route fit.

6. The method according to claim 1, characterized in that, The real-time monitoring of user feedback and behavior data, and the optimization of the contextualized strategy using the real-time monitoring of user feedback and behavior data, include: Using the aforementioned user feedback and behavioral data, and based on the time-series analysis of the behavioral data, sudden changes in the user's interaction time on specific types of content are identified. When the dwell time decreases beyond a set threshold, it is determined that the point of interest has shifted. Detect changes in interest points based on behavioral data. If the user's interaction time on a specific type of content decreases by more than a set threshold, it is determined that the user's interest points have shifted. Otherwise, it is determined that the point of interest has not shifted; Using the results of user feedback and interest point change detection, a reinforcement learning algorithm is employed to dynamically adjust the type weights of recommended content and optimize the contextualization strategy.

7. The method according to claim 1, characterized in that, The collected multi-dimensional user data includes: Data is collected through terminal devices, including mobile phones, smart bracelets, and smartwatches; the user's real-time behavioral data includes steps taken, heart rate, time spent at scenic spots, and frequency of page interactions. The user preference data includes interest tags, historical likes, and search keywords; The relevant historical data includes historical exploration routes, learning content completion records, and past itinerary arrangements.

8. The method according to claim 5, characterized in that, The deep learning model includes: In the natural exploration scenario, the LSTM model is used to predict the physical fitness of the current route based on the user's historical physical exertion curve and route adaptation record. In parent-child education scenarios, a CNN model is used to perform image feature mapping on the features in the user profile and generate appropriate graphic and text learning content. In tourism scenarios, the Transformer model is used to generate time-series itinerary recommendation sequences by combining user preferences with real-time weather and attraction crowding data.

Images