A multi-dimensional data-based interest level authentication method and system

Abstract

The application discloses a kind of interest level authentication method and system based on multidimensional data, it is related to big data interest grading and trusted evidence technology field, to solve the problems of single dimension of existing interest authentication mode, easy to be interfered by brushing, lack of dynamic optimization and low authentication result credibility.The method first complies with the collection of user multidimensional behavior data, and extracts real interest characteristics by time series behavior entropy and mutation coefficient check;Second, using the main auxiliary interest ladder coupling mechanism and multi-scene AI video evaluation model, a double-path authentication system is constructed to calculate the initial authentication score;With user medium and long term retention rate as objective function, the monthly collaborative iterative optimization of scoring and authentication weight is carried out by gradient descent algorithm;Finally, according to the final interest authentication level after optimization, the hierarchical lightweight blockchain evidence storage strategy is executed, and the user data of different levels is hashed and stored in the alliance chain side chain or main chain, to realize the accurate quantification, dynamic evaluation and trusted right of interest ability.

Description

Technical Field

[0001] This invention belongs to the cross-technical fields of big data user behavior mining, multimodal information processing, adaptive algorithm optimization and credible qualification storage. Specifically, it relates to an interest level certification method and system based on multi-dimensional data, which can be applied to user interest rating, growth ladder management and professional ability qualification certification scenarios in interest communities of all categories such as sports, handicrafts, outdoor activities and cultural entertainment. It can be deployed in APP, mini-program, offline venue check-in management system and event management platform. Background Technology

[0002] The current industry-wide interest level certification system generally relies on the traditional model of manual review and rating, supplemented by a small amount of single-dimensional user click and browsing data as a reference. This system has several shortcomings: First, the data source is too narrow, relying only on surface-level data such as explicit clicks and check-ins, failing to capture implicit multimodal behaviors such as viewing time, interaction preferences, and practical exercises. This results in a one-sided rating that cannot objectively reflect the user's true depth of interest and practical ability. Second, the manual review model is heavily influenced by the reviewers' subjective experience, resulting in poor standardization. Furthermore, the platform is prone to manipulation through fake orders and batch scripts. The fraudulent use of card data is rampant, and the existing single verification rules cannot identify machine-generated data. Users who manipulate scores can easily bypass certification levels, leading to a continuous decline in the credibility of certification. Third, the current rating and scoring rules and weights are all manually fixed before the system goes live. After launch, they cannot be dynamically and adaptively adjusted according to changes in user behavior and product category ecosystems. The rules have been rigid for a long time, resulting in frequent mismatches in rating, such as high scores with low retention and low scores with high capabilities. Fourth, there is no standardized evidence storage scheme for user certification qualifications. Paper certificates and local electronic certificates on the platform are easily tampered with or lost, and third-party institutions lack credible traceability evidence when verifying qualifications.

[0003] While a few digital interest-based rating platforms on the market have eliminated all manual review, they still use a fixed scoring formula plus a single-dimensional data statistics scheme. This still cannot solve the industry pain points of anti-fraud, dynamic parameter adjustment, and credible qualification confirmation. There is a lack of a full-chain intelligent authentication solution that integrates multi-source data anti-fraud, primary and secondary interest coupled scoring, AI video practical assessment, algorithm self-iteration optimization, and blockchain graded evidence storage. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this invention proposes an interest level authentication method, system, electronic device, and computer-readable storage medium based on multi-dimensional data.

[0005] The technical solution adopted in this invention is as follows: The first aspect of this invention provides a method for interest level authentication based on multi-dimensional data, comprising the following steps: S1. With the user's privacy authorization obtained, collect the user's multi-dimensional explicit behavior data and implicit multimodal behavior data to construct the original behavior dataset; if the user's authorization is not obtained, only collect explicit behavior data and set the implicit behavior correction coefficient to zero to construct the restricted original behavior dataset. S2. Receive the original behavior dataset or restricted original behavior dataset, perform weekly granular time-series slicing processing on it, calculate the user behavior entropy value based on preset constraints and combine it with time-series mutation coefficient verification, filter out invalid abnormal data in the dataset, and extract the user's real and effective interest behavior feature vector. S3. Based on the user's real and valid interest behavior feature vector, the effective auxiliary interest ladder is screened through the pre-constructed category capability association coupling coefficient library, and the feature vector is mapped to the corresponding ladder base score. The score of the effective auxiliary interest ladder is converted and incorporated into the main interest ladder using the main and auxiliary ladder coupling scoring rules, and user personalized interest ladder level data containing ladder base score is generated. S4. Obtain the user's personalized interest ladder level data, build a dual-path authentication access system, and evaluate the user's uploaded practical video based on video scene tags by adaptively switching multiple AI key point weight models to obtain the practical assessment score. The ladder base score is then used as a priori parameter to be introduced into the evaluation model to calculate the initial authentication score. S5. Collect user retention data after the initial authentication score. Using the maximization of the medium- and long-term retention rate after user authentication as the objective function, the gradient descent algorithm is used to perform monthly multi-parameter collaborative self-evolutionary iterative optimization on the ladder scoring weight in step S3 and the authentication scoring weight in step S4, and output the optimized weight parameter set. S6. Apply the optimized set of weight parameters to recalculate the user's final interest authentication level, and execute a layered lightweight blockchain evidence storage strategy based on the final interest authentication level. Hash and store low-level user data in the consortium blockchain sidechain, and hash and upload high-level user and authentication qualification data to the consortium blockchain main chain.

[0006] A second aspect of the present invention provides an interest level authentication system based on multi-dimensional data, for executing the aforementioned interest level authentication method based on multi-dimensional data, comprising: The compliant multi-source heterogeneous data collection module is used to collect multi-dimensional explicit behavioral data and implicit multimodal behavioral data with the user's privacy authorization, and to build the original behavioral dataset. The temporal behavior entropy feature extraction module is used to receive the original behavior dataset, perform weekly granular temporal slicing processing, calculate the user behavior entropy value and combine it with the temporal mutation coefficient verification to extract the user's real and effective interest behavior feature vector. The coupled and linked self-evolutionary algorithm engine module includes a main and auxiliary ladder division unit, a multi-ladder coupled scoring unit, a multi-scenario AI video practical evaluation unit, and a multi-parameter collaborative iterative optimization unit. It is used to map the feature vector to the ladder basic score, realize cross-category interest score linkage conversion, video evaluation scoring, and perform collaborative self-evolutionary iterative optimization on the ladder scoring weight and authentication scoring weight based on the retained data. The layered blockchain rights confirmation application module is used to calculate the final interest authentication level using the optimized set of weight parameters, and to execute a layered evidence storage strategy based on the level.

[0007] A third aspect of the present invention provides an electronic device comprising a storage module, a processing module, and a transceiver module connected in sequence for communication, wherein the storage module is used to store a computer program, the transceiver module is used to send and receive messages, and the processing module is used to read the computer program and execute the steps of the interest level authentication method based on multi-dimensional data.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the interest level authentication method based on multi-dimensional data.

[0009] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This invention breaks through the inherent architecture of traditional manual rating and single data statistics, and innovatively integrates six core technologies: privacy-graded data collection, dual anti-cheating verification of behavioral entropy and temporal mutation, main and auxiliary interest-coupled scoring and cross-border promotion, multi-scenario adaptive AI video practical evaluation, monthly autonomous parameter tuning of gradient descent with user retention as the optimization goal, and hierarchical blockchain hash storage. It completes the technical upgrade from the entire chain of data source, anti-cheating rules, scoring logic, practical assessment, parameter iteration, and qualification storage, and solves the pain points of existing technologies such as single dimension, rampant data fraud, rigid rules, distorted ratings, and easy tampering of certificates. Under the premise of compliant protection of user privacy, it realizes automated and accurate evaluation of interest level, dynamic adaptive iteration of rules, and credible on-chain confirmation of certification qualifications. Attached Figure Description

[0010] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 A flowchart of the method provided in Embodiment 1 of this application; Figure 2 This is a structural block diagram of the system provided in Embodiment 2 of this application; Figure 3 This is a schematic diagram of the structure of the computer device provided in Embodiment 3 of this application. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0012] Example 1:

[0013] As shown in Figure 1, this embodiment provides an interest level authentication method based on multi-dimensional data. The entire method relies on the collaborative implementation of multiple technologies, including big data streaming computing, information entropy verification, multi-category score coupling, AI human key evaluation, gradient descent parameter self-optimization, and consortium blockchain hierarchical evidence storage. It is applicable to user leveling and qualification authentication on various interest community platforms such as sports, handicrafts, and entertainment, and includes the following steps: S1. With the user's privacy authorization obtained, collect multi-dimensional explicit behavioral data and implicit multimodal behavioral data from the user in compliance with regulations. Explicit behavioral data refers to directly quantifiable statistical data generated by the user's active operations, including check-in, course registration, content publishing, and event registration. Implicit multimodal behavioral data refers to potential preference data mined from audio and video and interactive behaviors, covering information such as effective viewing time, repeated playback of segments, practice frequency, and interactive emotional tendencies. The original behavioral dataset is constructed based on the full collection of data. If the user's privacy authorization is not obtained, the principle of minimum collection for personal information protection is strictly followed. Only explicit behavioral data within the scope of compliance is collected, and the implicit behavior correction coefficient in the formula is set to zero. All implicit data collection links are eliminated, and a restricted original behavioral dataset is constructed to achieve privacy hierarchical control from the source. S2. Receive the original behavior dataset or restricted original behavior dataset, perform weekly-granular time-series slicing on the full business data according to the natural week time caliber, split it into segmented data subsets with a single week as the unit, calculate the user behavior entropy value based on preset constraints and combine the time-series mutation coefficient dual verification rule to accurately identify cheating behaviors such as batch script brushing and short-term concentrated flooding, filter out invalid abnormal data in the dataset, and then perform data standardization and feature normalization processing to extract the user's real and effective interest behavior feature vector, and eliminate the interference caused by abnormal data to subsequent scoring; S3. Based on the user's real and valid interest behavior feature vector, call the pre-built and continuously maintained and updated category capability association coupling coefficient library, filter the effective auxiliary interest ladder according to the category association value, and convert the normalized feature vector into the corresponding ladder base score according to the preset score mapping rule. Using the main and auxiliary ladder coupling scoring rule, the score of the selected effective auxiliary interest ladder is converted into the main interest ladder score according to the coupling ratio. Combined with the differentiated promotion conditions, the user's personalized interest ladder level data containing the ladder base score is finally generated, realizing the cross-category interest capability value interoperability. S4. Obtain the user's personalized interest ladder level data, and build a dual-path authentication and access system consisting of historical behavior ladder scoring and real-scene video practical assessment. First, parse the metadata fields of the user's uploaded practical video to automatically identify video scene tags. Based on different scene tags, adaptively switch multiple sets of AI key point weight models. Rely on human posture key point recognition technology to evaluate the user's uploaded practical video frame by frame to obtain the practical assessment score. The ladder basic score generated in the previous step is used as a priori reference parameter and introduced into the score calculation logic of the evaluation model. After weighted fusion of the two types of scores, the initial authentication score is calculated. S5. Collect user retention data throughout the entire lifecycle after the initial authentication score is completed and the qualification assessment is completed. Take the maximization of the medium- and long-term retention rate after user authentication as the optimization objective function. Use the gradient descent algorithm to perform monthly multi-parameter collaborative self-evolutionary iterative optimization on the ladder scoring weight in step S3 and the authentication scoring weight in step S4. The optimization process constrains the fluctuation range of parameters to avoid the distortion of weight sudden changes. After the iteration converges, the optimized weight parameter set is output to realize the scoring rules autonomously iterate and upgrade with the platform user ecosystem. S6. Apply the optimized set of weight parameters, input the original scoring logic to recalculate the user's final interest certification level, and execute a layered lightweight blockchain evidence storage strategy based on the final interest certification level. Hash and store the L1-L3 low-level user data in the consortium blockchain sidechain to save main chain storage resources, and hash and upload the full data of L4-L5 high-level users and certification qualifications to the consortium blockchain main chain to complete the authoritative confirmation.

[0014] In step S2, the calculation of user behavior entropy based on preset constraints and verification using temporal abrupt change coefficients specifically includes: The behavioral entropy value is calculated using the following formula. :

[0015] in, For the first The percentage of frequency of a particular behavior out of the total frequency of behaviors. The total number of behavior categories. For fixed correction factors, This is a correction factor for implicit behavior. The process of filtering out invalid and abnormal data in the dataset specifically includes: when Hu is less than a preset judgment threshold, it indicates that the user behavior distribution is highly uniform and fixed, and the data corresponding to this user's period is judged as script-generated data and removed; when Hu is greater than or equal to the preset judgment threshold but the time-series mutation coefficient exceeds the preset increase benchmark, it is judged as short-term batch data injection and removed, and special users such as contracted coaches and professional players are included in the high-entropy user whitelist, and the data removal operation is exempted when the threshold is triggered.

[0016] In step S3, the scores of the effective auxiliary interest ladders are converted and incorporated into the main interest ladder using the main and auxiliary ladder coupling scoring rules. Specifically, this includes: pre-statistically calculating the ability correlation between all categories of interests, setting the category correlation coupling coefficient Kn and the admission threshold K. If the Kn of the category to which the auxiliary interest ladder belongs is greater than or equal to the K threshold of the category to which the main interest ladder belongs, it means that the two types of interest abilities are correlated, and then the auxiliary interest ladder is marked as an effective auxiliary interest ladder. When calculating the total score of the main interest ladder, a weighted score from the effective secondary interest ladder is introduced. The calculation formula is as follows:

[0017] in, The main ladder base score To effectively assist in ranking up, and To initialize the weight parameters.

[0018] In detail, step S3 generates personalized interest-based ladder level data for users, including basic ladder scores, and also includes a parameterized differentiated promotion mechanism: When a user's main interest ladder score does not reach the threshold score for this level, but the score gap is less than the preset ratio parameter Tgap, and the number of valid secondary interest ladders owned by the user is not less than the preset number and the level reaches the preset higher level, cross-border linkage promotion is triggered. It is not restricted by the main score shortfall, and the user is judged to be promoted and the corresponding ladder level data is generated.

[0019] In step S4, multiple AI keypoint weight models are adaptively switched based on video scene labels to evaluate the user-uploaded practical videos. Specifically, this includes: The FFmpeg tool is used to parse the metadata fields of the uploaded video and extract content such as notes, shooting scenes, and event logos to obtain video scene tags. If the video scene is labeled as "beginner practice", the first weight model is used, with 70% weight for action completeness and 30% weight for action standardization. If the video scene is tagged as "daily practical operation", the second weight model is adopted, with the action standardization weight at 60% and the action completeness weight at 40%. If the video scene is labeled as a live match, the third-weight model is used, with competitive performance weighted at 75% and basic movement completeness weighted at 25%.

[0020] In step S5, the gradient descent algorithm is used to perform monthly multi-parameter collaborative self-evolutionary iterative optimization on the ladder scoring weights in step S3 and the authentication scoring weights in step S4, specifically including: Starting from the date of issuance of the user certificate, user activity retention data is continuously collected for 30 natural days. Users with a retention rate greater than 40% are classified as high-quality samples, and users with a retention rate less than 10% are classified as low-quality samples. The scoring weights are then adjusted in reverse based on the sample distribution. Construct the optimization objective function:

[0021] in The weights are respectively for practical skills, theoretical skills, and other sub-items; The gradient descent algorithm is used to iteratively update all parameters, with the following constraint: the absolute value of the change in any parameter in a single iteration does not exceed 5%, to prevent drastic parameter oscillations from causing an imbalance in the scoring system.

[0022] In step S6, a layered lightweight blockchain evidence storage strategy is implemented based on the final interest authentication level, specifically including: For lower-level users at levels L1 to L3, their weekly growth summary data is encrypted with SHA to generate a hash value and then stored in the consortium blockchain sidechain. The original detailed data is stored in the platform's local database, reducing on-chain data storage costs. For advanced users at levels L4 to L5 and all users who have obtained official certification, their full information hash value will be uploaded to the consortium blockchain main chain, and a unique on-chain confirmation number will be generated after consensus among the on-chain nodes. The platform pre-deploys automatically executed verification smart contracts. Subsequently, it responds to verification requests from enterprises, universities, and third-party institutions through the deployed smart contracts, comparing the hash value of externally input data with the hash value stored on the chain to verify data integrity, thereby achieving tamper-proof qualifications and reliable traceability.

[0023] Example 2:

[0024] like Figure 2 As shown, this embodiment provides an interest level authentication system based on multi-dimensional data, used to execute the interest level authentication method based on multi-dimensional data, including: The compliant multi-source heterogeneous data acquisition module is used to collect multi-dimensional explicit behavioral data and implicit multimodal behavioral data under the premise of obtaining user privacy authorization, and to build the original behavioral dataset. This module integrates SDK tracking, API interface docking and IoT device data access functions, and supports automatic switching to the "minimum data acquisition mode" for unauthorized users.

[0025] The temporal behavior entropy feature extraction module is used to receive the original behavior dataset, perform weekly temporal slicing processing, calculate the user behavior entropy value and combine it with temporal mutation coefficient verification to extract the user's real and effective interest behavior feature vector; the module has a built-in sliding window algorithm, which can monitor the entropy value change of the data stream in real time and automatically intercept abnormal traffic.

[0026] The coupled and interconnected self-evolving algorithm engine module includes a primary and secondary ladder partitioning unit, a multi-ladder coupled scoring unit, a multi-scenario AI video practice evaluation unit, and a multi-parameter collaborative iterative optimization unit. This unit maps the feature vectors to basic ladder scores, enabling cross-category interest score linkage conversion and video evaluation scoring. It also performs collaborative self-evolving iterative optimization of ladder scoring weights and authentication scoring weights based on retained data. The AI ​​video practice evaluation unit is pre-configured with a pose estimation network (such as OpenPose or MoveNet), which can dynamically load different weight configuration files based on scene labels.

[0027] The layered blockchain rights confirmation application module is used to calculate the final interest authentication level using the optimized set of weight parameters, and to execute a layered evidence storage strategy based on the level. This module encapsulates the blockchain SDK, which can automatically route data to the main chain or side chain nodes according to the user level, and provides standardized API interfaces for third parties to call for verification.

[0028] Example 3:

[0029] like Figure 3 As shown, this embodiment provides an electronic device, including a storage module, a processing module, and a transceiver module that are sequentially and communicatively connected. The storage module is used to store a computer program, the transceiver module is used to send and receive messages, and the processing module is used to read the computer program and execute the steps of the interest level authentication method based on multi-dimensional data.

[0030] Specifically, the electronic device can manifest as a high-performance server cluster or a cloud computing instance. The storage module includes high-speed random access memory (RAM) for caching real-time behavioral data, and non-volatile memory (such as SSD) for storing historical behavioral logs, model parameters, and blockchain ledger data. The processing module includes a central processing unit (CPU) and a graphics processing unit (GPU), where the CPU is responsible for logic control, data preprocessing, and entropy calculation, and the GPU accelerates matrix operations for AI video keypoint detection and gradient descent algorithms. The transceiver module includes a gigabit Ethernet interface or a 5G communication module for high-concurrency reception of user-uploaded video streams and behavioral logs, and for sending authentication results and blockchain evidence hashes back to the user terminal.

[0031] Example 4:

[0032] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the interest level authentication method based on multi-dimensional data.

[0033] Specifically, the computer-readable storage medium can be a read-only memory (ROM), a programmable read-only memory (PROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), a disk, or an optical disk, etc. The computer program is written using a modular architecture, including data cleaning subroutines, feature engineering subroutines, model inference subroutines, and blockchain interaction subroutines. When the computer program is loaded into memory and executed by the processor, it enables the computing device to complete the cleaning and feature extraction of massive amounts of user behavior data within nanoseconds, complete AI video evaluation and score calculation within seconds, and complete the self-evolutionary update of weight parameters and blockchain consensus ledger for the entire network within minutes, thereby ensuring the real-time performance and consistency of the system under high load.

[0034] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for interest level authentication based on multi-dimensional data, characterized in that, Including the following steps: S1. With the user's privacy authorization obtained, collect the user's multi-dimensional explicit behavioral data and implicit multimodal behavioral data to construct the original behavioral dataset; If user authorization is not obtained, only explicit behavior data will be collected and the implicit behavior correction coefficient will be set to zero to construct a restricted raw behavior dataset; S2. Receive the original behavior dataset or restricted original behavior dataset, perform weekly granular time-series slicing processing on it, calculate the user behavior entropy value based on preset constraints and combine it with time-series mutation coefficient verification, filter out invalid abnormal data in the dataset, and extract the user's real and effective interest behavior feature vector. S3. Based on the user's real and valid interest behavior feature vector, the effective auxiliary interest ladder is screened through the pre-constructed category capability association coupling coefficient library, and the feature vector is mapped to the corresponding ladder base score. The score of the effective auxiliary interest ladder is converted and incorporated into the main interest ladder using the main and auxiliary ladder coupling scoring rules, and user personalized interest ladder level data containing ladder base score is generated. S4. Obtain the user's personalized interest ladder level data, build a dual-path authentication access system, and evaluate the user's uploaded practical video based on video scene tags by adaptively switching multiple AI key point weight models to obtain the practical assessment score. The ladder base score is then used as a priori parameter to be introduced into the evaluation model to calculate the initial authentication score. S5. Collect user retention data after the initial authentication score. Using the maximization of the medium- and long-term retention rate after user authentication as the objective function, the gradient descent algorithm is used to perform monthly multi-parameter collaborative self-evolutionary iterative optimization on the ladder scoring weight in step S3 and the authentication scoring weight in step S4, and output the optimized weight parameter set. S6. Apply the optimized set of weight parameters to recalculate the user's final interest authentication level, and execute a layered lightweight blockchain evidence storage strategy based on the final interest authentication level. Hash and store low-level user data in the consortium blockchain sidechain, and hash and upload high-level user and authentication qualification data to the consortium blockchain main chain.

2. The method according to claim 1, characterized in that, In step S2, the calculation of user behavior entropy based on preset constraints and verification using temporal abrupt change coefficients specifically includes: The behavioral entropy value is calculated using the following formula. : in, For the first The percentage of frequency of a particular behavior out of the total frequency of behaviors. The total number of behavior categories. For fixed correction factors, This is a correction factor for implicit behavior. The process of filtering out invalid and abnormal data from the dataset specifically includes: when When the data is less than the preset threshold, it is determined to be script-generated data and is removed; when If the data is greater than or equal to the preset threshold but the temporal mutation coefficient exceeds the preset increase benchmark, it is judged as abnormal data and removed, unless the user exists in the high-entropy user whitelist.

3. The method according to claim 1, characterized in that, In step S3, the scores of effective secondary interest ladders are converted and incorporated into the main interest ladder using the main and secondary ladder coupling scoring rules. Specifically, this includes setting the category association coupling coefficient. and admission threshold If the category of the secondary interest ladder is the same as the category of the primary interest ladder... If so, then mark the auxiliary interest ladder as a valid auxiliary interest ladder; When calculating the total score of the main interest ladder, a weighted score from the effective secondary interest ladder is introduced. The calculation formula is as follows: in, The main ladder base score To effectively assist in ranking up, and To initialize the weight parameters.

4. The method according to claim 3, characterized in that, In step S3, personalized interest-based ladder ranking data containing basic ladder scores is generated, and a parameterized differentiated promotion mechanism is also included: When a user's main interest ranking score does not reach the threshold score for this level, but the score gap is less than the preset ratio parameter. When a user has at least a preset number of valid secondary interest ladders and reaches a preset high-level level, cross-domain linkage promotion is triggered, the user is promoted, and corresponding ladder level data is generated.

5. The method according to claim 1, characterized in that, In step S4, multiple AI keypoint weight models are adaptively switched based on video scene labels to evaluate the user-uploaded practical videos. Specifically, this includes: Parse the metadata of the uploaded video to obtain video scene tags; If the video scene is labeled as "beginner practice", the first weight model is used, with 70% weight for action completeness and 30% weight for action standardization. If the video scene is tagged as "daily practical operation", the second weight model is adopted, with the action standardization weight at 60% and the action completeness weight at 40%. If the video scene is labeled as a live match, the third-weight model is used, with competitive performance weighted at 75% and basic movement completeness weighted at 25%.

6. The method according to claim 1, characterized in that, In step S5, the gradient descent algorithm is used to perform monthly multi-parameter collaborative self-evolutionary iterative optimization on the ladder scoring weights in step S3 and the authentication scoring weights in step S4. Specifically, this includes: The retention rate of certified users over 30 consecutive calendar days from the date of certificate issuance is statistically analyzed. Users with a retention rate greater than 40% are classified as high-quality samples, and users with a retention rate less than 10% are classified as low-quality samples. Construct the optimization objective function: in The weights are respectively for practical skills, theoretical skills, and other sub-items; The gradient descent algorithm is used to update the parameters, with the constraint that the absolute value of the change in any parameter in a single iteration does not exceed 5%.

7. The method according to claim 1, characterized in that, In step S6, a layered lightweight blockchain evidence storage strategy is executed based on the final interest authentication level, specifically including: For lower-level users at levels L1 to L3, only the hash value of their weekly growth summary data is encrypted and stored in the consortium blockchain sidechain, while the original detailed data is stored locally. For advanced users at levels L4 to L5 and all users who have obtained certification, their full information hash value is uploaded to the consortium blockchain main chain to generate a unique on-chain confirmation number. The deployed smart contract responds to third-party verification requests, comparing the hash value of the input data with the hash value stored on the chain to verify data integrity.

8. An interest-level authentication system based on multi-dimensional data, characterized in that, For performing the method according to any one of claims 1 to 7, comprising: The compliant multi-source heterogeneous data collection module is used to collect multi-dimensional explicit behavioral data and implicit multimodal behavioral data with the user's privacy authorization, and to build the original behavioral dataset. The temporal behavior entropy feature extraction module is used to receive the original behavior dataset, perform weekly granular temporal slicing processing, calculate the user behavior entropy value and combine it with the temporal mutation coefficient verification to extract the user's real and effective interest behavior feature vector. The coupled and linked self-evolutionary algorithm engine module includes a main and auxiliary ladder division unit, a multi-ladder coupled scoring unit, a multi-scenario AI video practical evaluation unit, and a multi-parameter collaborative iterative optimization unit. It is used to map the feature vector to the ladder basic score, realize cross-category interest score linkage conversion, video evaluation scoring, and perform collaborative self-evolutionary iterative optimization on the ladder scoring weight and authentication scoring weight based on the retained data. The layered blockchain rights confirmation application module is used to calculate the final interest authentication level using the optimized set of weight parameters, and to execute a layered evidence storage strategy based on the level.

9. An electronic device, characterized in that, The device includes a storage module, a processing module, and a transceiver module that are sequentially connected in communication. The storage module is used to store a computer program, the transceiver module is used to send and receive messages, and the processing module is used to read the computer program and execute the steps of the interest level authentication method based on multi-dimensional data as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the interest level authentication method based on multi-dimensional data as described in any one of claims 1 to 7.

Images