Label management method, device and equipment and storage medium

Abstract

The application belongs to the technical field of data processing, and discloses a label management method, device, equipment and storage medium, the label management method comprises the following steps: based on a preset value decay model, time decay weighting is performed on behavior index data of each dimension respectively, and the comprehensive value score of a preset label is evaluated based on the weighted data; according to the comparison result of the comprehensive value score and the dynamic threshold, the temperature state of the preset label is determined; when the temperature state is a cold label, an automatic archiving operation is triggered, and the data associated with the preset label is migrated to a low-cost storage medium; when a calling request for a target cold label is detected, an asynchronous wake-up process is started to reconstruct the label result of the target cold label, and the comprehensive value score of the target cold label is corrected. The application can be applied to financial technology, medical health and other business management systems, and solves the technical problem that the label management method cannot effectively control system resource consumption under the premise of ensuring label availability.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology and is applied in the fields of financial technology and healthcare. In particular, it relates to a tag management method, device, equipment and storage medium. Background Technology

[0002] In data-driven enterprises, tagging platforms bear the core responsibility of transforming raw data into business-usable tags. As businesses grow, the number of tags on the platform often grows rapidly to hundreds, thousands, or even tens of thousands, and is widely used in scenarios such as audience segmentation, personalized recommendations, and marketing analytics.

[0003] In the life insurance sector, tagging platforms manage a large number of customer insight tags such as "potential clients for high-end senior living communities" and "high-net-worth churn warnings," supporting agents in business development and precision marketing. In the healthcare sector, tagging platforms carry patient risk stratification tags such as "high-risk patients with diabetic foot" and "high-risk patients requiring postoperative follow-up," assisting in clinical decision-making and chronic disease management.

[0004] However, with the continuous expansion of the number of tags, existing tag platforms face common operational challenges in lifecycle management. Many tags are used at varying frequencies after creation, with some tags remaining in a low-usage or even zero-usage state for extended periods. However, the platform lacks effective identification methods and can only indiscriminately allocate computing resources and high-speed storage space to all tags. Operations personnel typically rely on manual periodic inventory checks and experience to determine which tags can be decommissioned or migrated. This method is not only outdated but also prone to misjudgment, resulting in insufficient resource guarantees for valuable tags while excessive resource consumption by low-value tags.

[0005] Furthermore, when tags already in a low-activity state are occasionally invoked by business applications, a dilemma arises in how to respond: keeping them fully online continuously consumes resources, while taking them offline completely prevents rapid recovery upon invocation, impacting business continuity. These problems become more pronounced in systems with larger tag scales and more complex business scenarios, resulting in high total cost of ownership for the tag platform and hindering operational efficiency. Therefore, there is an urgent need for a method that can automate and finely manage the tag lifecycle to effectively control system resource consumption while ensuring tag availability. Summary of the Invention

[0006] This invention provides a tag management method, apparatus, device, and storage medium, which can solve the technical problem that existing tag management methods cannot effectively control system resource consumption while ensuring tag availability.

[0007] In a first aspect, the present invention provides a label management method, comprising: Collect behavioral indicator data from multiple dimensions associated with preset tags; Based on a preset value decay model, the behavioral indicator data of each dimension are weighted by time decay, and the comprehensive value score of the preset label is evaluated based on the weighted data. The temperature status of the preset label is determined based on the comparison result between the comprehensive value score and the dynamic threshold. When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. When a call request for a target cold tag is detected, an asynchronous wake-up process is initiated to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

[0008] In a second aspect, the present invention provides a label management device, comprising: The data acquisition module is used to collect behavioral indicator data from multiple dimensions associated with preset tags; The value assessment module is used to perform time decay weighting on the behavioral indicator data of each dimension based on a preset value decay model, and to evaluate the comprehensive value score of the preset label based on the weighted data. The status determination module is used to determine the temperature status of the preset label based on the comparison result between the comprehensive value score and the dynamic threshold. The data processing module is used to trigger an automatic archiving operation when the temperature status is determined to be a cold tag. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. The score correction module is used to initiate an asynchronous wake-up process when a call request for a target cold tag is detected, in order to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

[0009] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described tag management method.

[0010] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described tag management method.

[0011] Understandably, the present invention, through the steps of "collecting behavioral indicator data from multiple dimensions associated with preset tags" and "applying time-decay weighting to the behavioral indicator data of each dimension based on a preset value decay model, and evaluating the comprehensive value score of the preset tag based on the weighted data," transforms the originally unquantifiable "tag value" into a calculable and comparable comprehensive value score. Time-decay weighting makes recent behavior contribute more significantly to the value, and the score dynamically reflects the tag's current activity level, rather than relying solely on static creation time or historical cumulative data. This step provides a quantitative basis for subsequent automated decision-making.

[0012] Furthermore, the present invention, through the step of "determining the temperature state of a preset label based on the comparison result of the comprehensive value score and the dynamic threshold," maps continuous value scores to discrete temperature states, establishing a decision logic of "high value corresponding to high resource guarantee, and low value corresponding to low resource consumption." The introduction of the dynamic threshold enables the judgment criteria to adaptively adjust with the overall resource pressure of the system, tightening the criteria when resources are scarce and relaxing the criteria when resources are abundant, ensuring that resource allocation always matches the actual carrying capacity of the system.

[0013] Furthermore, the present invention implements the temperature status decision into a substantial resource recovery action by stepping on the following: "When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered, including suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium." Suspending the calculation task directly releases CPU and memory resources, while migrating the data to a low-cost storage medium releases expensive high-performance storage space, thereby achieving resource overhead control for low-value tags.

[0014] Furthermore, the present invention addresses the potential availability degradation issue following resource reclamation by initiating an asynchronous wake-up process upon detecting a call request for a target cold tag, thereby reconstructing the tag result of the target cold tag and correcting its overall value score. Asynchronous wake-up ensures that archived cold tags can still be reconstructed and have results returned when needed, rather than becoming completely unusable. Correcting the overall value score prevents woken-up tags from immediately triggering archiving again due to historically low scores, avoiding the jarring effect of wake-up immediately followed by archiving and providing tags with a reasonable period of active observation.

[0015] In summary, the present invention first quantifies the value of tags into comparable scores, and then automatically performs differentiated allocation and recycling of resources based on these scores. This further ensures that tags whose resources have been recycled can still be activated for service when needed. This closed loop allows the system to concentrate resources on a few high-value tags most of the time, while simultaneously recycling resources from a large number of low-value tags, significantly reducing overall resource overhead. Furthermore, when low-value tags are occasionally invoked by business applications, an asynchronous wake-up mechanism can be used to guarantee their availability, thus achieving the goal of "effectively controlling system resource overhead while ensuring tag availability." Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a label management method according to an embodiment of the present invention.

[0018] Figure 2 yes Figure 1 A flowchart of step S130.

[0019] Figure 3 This is another flowchart illustrating the label management method in one embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of a tag management device according to an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention.

[0022] Figure 6 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Figure 1 A flowchart of the tag management method provided in the embodiments of the present invention is shown below. Figure 1As shown, the label management method provided in this embodiment of the invention includes the following steps.

[0025] Step S110: Collect behavioral indicator data from multiple dimensions associated with preset tags; Specifically, in this step, we can first define multiple observation dimensions to measure the value of tags, which can at least cover four aspects: tag usage popularity, business effectiveness, data freshness, and user feedback. Specifically, we can obtain the call frequency of each preset tag in real-time or near real-time through the tag service gateway, scheduling engine logs, and event tracking data collection system; by connecting to the business effect attribution platform, we can collect indicators such as conversion rate and response rate improvement brought about by the tag's effect on specific business activities; through the metadata management module, we can continuously monitor the last update time and update frequency of the upstream data tables or data streams on which the tag depends, forming data source timeliness indicators; simultaneously, by collecting user feedback events such as likes, error corrections, and effectiveness scores from the front-end application, we can obtain subjective evaluation dimensions of tag quality. The collected indicators are written to the indicator storage in the form of time-series events, with each indicator record bearing a timestamp, providing a basis for subsequent decay calculations.

[0026] As a concrete example in the life insurance field, we can take the tag "potential customers of high-end health and wellness communities" as an example. We can collect data on the number of times the tag has been used by agents' business development apps and marketing platforms in the past 30 days, the customer participation conversion rate improvement value of health and wellness community experience invitation activities driven by the tag, the most recent update interval of the customer's investable asset data source and health assessment data, and the accurate or inaccurate feedback records given by agents regarding the tag.

[0027] As another concrete example, in the healthcare field, taking the label "high-risk patients with diabetic foot" as an example, we collect data on the frequency with which the label is used by the chronic disease management platform, the increase in patient foot care compliance rate relative to the baseline period after the application of the label, the data of the associated blood glucose monitoring devices, the reporting frequency and latest upload time of electronic medical records, and the effectiveness scores or error correction records of endocrinologists for the risk grading label.

[0028] In some embodiments of the present invention, the multiple dimensions of behavioral indicator data include indicator data representing the frequency of tag calls, indicator data representing the contribution improvement generated by the tag in business activities, indicator data representing the update status of the upstream data source on which the tag depends, and indicator data representing the user's feedback behavior on the tag.

[0029] Specifically, in this embodiment, for behavioral indicator data, the first dimension represents the indicator data of tag call frequency. A call log collection component is deployed at the tag service gateway or unified query interface layer to record each tag query, filtering, batch export, and other call behaviors. Each log entry includes the tag identifier, the caller application identifier, and the call timestamp. The collection component sends the logs to a message queue asynchronously, where a stream processing task aggregates and calculates them according to a preset time window, generating the total number of calls and the number of unique callers for each tag within the window. The second dimension represents the indicator data of the contribution improvement generated by the tag in business activities. The tag usage is associated with the effect of downstream business activities through a business attribution system. When a tag is applied to a specific business activity, the association between the tag and the activity identifier (ID) is recorded. After the activity ends, the improvement in core business indicators for the target group applying the tag compared to the control group that did not apply the tag is calculated. This improvement value is expressed as an absolute percentage increase or a relative percentage increase. The third dimension represents the indicator data of the update status of the upstream data source on which the tag depends. The tag management system maintains a dependency mapping table between each tag and its upstream data source. The metadata collection module periodically accesses the metadata interface or system tables of the upstream data source to check the latest partition creation time or the latest offset submission time of each dependent data table, calculates the interval between the current time and the latest update time, and statistically analyzes the update frequency within the evaluation period to reflect the freshness of the tag's underlying data. The fourth dimension consists of metrics representing user feedback behavior on tags. A feedback component is embedded in the tag user interface, allowing users to evaluate or correct the accuracy and effectiveness of tag results. Collected feedback events include tag identifiers, feedback types, ratings or correction markers, and feedback timestamps. Feedback data is cleaned and deduplicated before being imported into the metrics storage.

[0030] As a concrete example in the life insurance field, taking the tag "potential customers of high-end senior living communities" as an example, the first dimension collects the total number of times this tag has been used by agents' sales apps and marketing automation platforms in the past 30 days, and counts the number of branches and agents involved in the independent calls. The second dimension collects the increase in invitation response rate brought about by this tag in a senior living community experience invitation campaign, that is, the absolute percentage point increase in response rate of the customer group selected by the tag compared with the randomly selected control group. The third dimension collects the interval between the latest partition creation time and the current time of the customer investable asset data table on which the tag is based, as well as the update frequency of the data table in the past 30 days. The fourth dimension collects the agent's feedback records of "accurate / inaccurate" for the tag and the effectiveness rating.

[0031] As another concrete example in the healthcare field, taking the label "high-risk diabetic foot patient" as an example, the first dimension collects the total number of times the label has been accessed by the chronic disease management platform and clinic support system in the past 30 days, and counts the number of departments and doctors covered by the accessors. The second dimension collects the percentage point increase in foot care compliance rate of the labeled patient group compared to the routine follow-up group after the label is applied to the intelligent follow-up grouping strategy for chronic disease management. The third dimension collects the interval between the most recent offset submission time and the current time of the real-time blood glucose monitoring data stream on which the label depends, as well as the update frequency of the glycated hemoglobin test record table in the electronic medical record database. The fourth dimension collects the "effectiveness score" and error correction records of the risk classification label by endocrinologists and vascular surgeons.

[0032] Understandably, this embodiment constructs a comprehensive and multi-dimensional tag value measurement system by using behavioral indicator data collected from four dimensions—tag call frequency, business activity contribution improvement, upstream data source update status, and user feedback behavior—as input for value assessment. The call frequency dimension captures the business usage intensity of the tag; the contribution improvement dimension measures the actual driving effect of the tag on business goals; the data source update status dimension reflects the timeliness and reliability of the tag's underlying data; and the user feedback dimension introduces direct human judgment on tag quality based on experience. These four types of indicators complement each other from the perspectives of usage intensity, business value, data freshness, and quality recognition, enabling the subsequent comprehensive value score assessment to more completely and accurately reflect the true vitality of a tag in the current system. This avoids evaluation biases that may arise from relying on a single dimension (such as only looking at the number of calls), laying a reliable data foundation for refined management of the tag lifecycle.

[0033] Step S120: Based on the preset value decay model, the behavioral indicator data of each dimension are weighted by time decay, and the comprehensive value score of the preset label is evaluated based on the weighted data. Specifically, in this step, the core idea of ​​the preset value decay model is that more recently occurring behaviors contribute more to the tag value. An exponential decay function can be used, with a configurable half-life parameter (e.g., 7 days or 14 days). For each behavioral indicator event collected in S110, the decay weight is calculated using the interval from its occurrence time to the current evaluation time as the independent variable. For example, the first... Weight of each event ,in The score is determined by the half-life. Events under each dimension are weighted and summed separately using this method to obtain a dimensional score for each dimension. Subsequently, a preset fusion algorithm (such as weighted summation, logistic regression, or normalized linear combination) is used to aggregate the dimensional scores of each dimension into a comprehensive value score, with the score range mapped to the interval [0,100] or [0,1]. The entire evaluation process is driven by a scheduled task, with a batch recalculation of all active tags daily or hourly.

[0034] As a specific example, in the life insurance field, for the tag of potential customers of high-end health and wellness communities, a recent (e.g., 2 days ago) high-frequency call event receives a weight close to 0.9, while a feedback behavior 25 days ago has a weight that decays to less than 0.2; after weighted fusion calculation, the comprehensive value score of this tag may be 52 points.

[0035] As another concrete example, in the field of healthcare, in the assessment of high-risk patients with diabetic foot, the high scores given by doctors within the most recent day have a very high weighting for effectiveness evaluation, while the large number of calls from a month ago contribute little due to decay, and the final calculated comprehensive value score is 48 points.

[0036] Step S130: Determine the temperature status of the preset label based on the comparison result of the comprehensive value score and the dynamic threshold. Specifically, in this step, a dynamic threshold can be maintained. This threshold is not a fixed constant but is linked to the overall resource pressure and operational status of the tag management system. The dynamic threshold calculation module periodically collects indicators such as the cluster's CPU utilization, memory usage, and the backlog of offline task queues, and generates a real-time threshold according to a preset mapping function (such as a piecewise linear function). Whenever the current comprehensive value score of a tag is calculated in S120, it is compared with the current dynamic threshold: if the comprehensive value score is equal to or higher than the dynamic threshold, the tag is marked as a hot tag; if the comprehensive value score is lower than the dynamic threshold, it is marked as a cold tag. The status change record is updated in the tag metadata storage for use by the scheduling and storage engine. This mechanism ensures that the cold tag identification threshold is automatically raised when the system is busy, freeing up more resources.

[0037] As a specific example, in the life insurance sector, during the peak sales season, system computing resources were strained and task queues were backlogged, causing the dynamic threshold to automatically increase from the usual 30 points to 55 points. The "potential customer for high-end senior living communities" tag, which originally had a comprehensive value score of 52 points, was therefore deemed a "cold tag."

[0038] As another concrete example, in the healthcare field, when the offline computing cluster is saturated, the dynamic threshold increases from 25 to 45 points, causing the label of a high-risk diabetic foot patient with an original score of 40 to enter a cold label state.

[0039] In some embodiments of the present invention, such as Figure 2 As shown, step S130 includes the following steps.

[0040] Step S131: When the comprehensive value score is higher than the first threshold, it is determined to be a hot tag; Specifically, in this step, the currently effective first threshold can be read from the tag metadata storage. This first threshold is not a fixed constant, but a value dynamically calculated and periodically refreshed by the threshold management module according to preset rules based on the overall real-time resource utilization of the tag management system (such as CPU load, memory usage, task queue depth, etc.). After obtaining the latest comprehensive value score of the tag to be evaluated, the tag status determination service performs a greater than comparison operation. If the score is strictly greater than the first threshold, the temperature status field of the tag is atomically updated to "hot tag". After the status update, a status change event is published through the message queue to notify the scheduling engine to maintain or resume the routine calculation task of the tag.

[0041] As a specific example, in the life insurance sector, during the operation of life insurance business, the current comprehensive value score of the tag "potential customer for high-end senior living communities" is 78 points. At this time, due to the system being in the peak of the "opening sales season" and high resource utilization, the first threshold is dynamically adjusted to 75 points. Because 78 points is higher than 75 points, the judgment logic is triggered, and the tag is judged as a "hot tag," and its corresponding daily customer segmentation calculation task continues to be executed as planned.

[0042] As another concrete example, in the healthcare field, within a healthcare data platform, the current comprehensive value score for the tag "high-risk diabetic foot patient" is 72 points. At the current moment, the system's stream processing cluster task slot utilization is approximately 60%, which is considered a medium load, and the first threshold is dynamically determined to be 70 points. Because 72 points is higher than 70 points, this tag is classified as a "hot tag," and its associated real-time risk assessment stream computing task remains active.

[0043] Step S132: When the comprehensive value score is lower than the second threshold, it is determined to be a cold label; The values ​​of the first threshold and the second threshold are dynamically adjusted according to the resource utilization rate of the tag management system.

[0044] Specifically, in this step, the system simultaneously maintains a second threshold below the first threshold, and the two together form a buffered judgment range. The value of the second threshold is also dynamically adjusted by the threshold management module based on system resource utilization. When the comprehensive value score of the tag to be evaluated is strictly below the second threshold, the temperature status field is atomically updated to "cold tag," and a status change event is published, triggering subsequent automatic archiving operations. It should be noted that the dynamic adjustment mechanism of the first and second thresholds is synchronized: when resource pressure increases, both thresholds are raised simultaneously, tightening the overall criteria for hot tags; when resource pressure decreases, both thresholds are lowered simultaneously, relaxing the criteria. This design ensures that the system can automatically balance resource consumption and tag service availability under different load conditions, while the difference between the two thresholds provides a stable range for tags to avoid frequent switching between hot and cold tags.

[0045] As a concrete example, in the life insurance sector, the comprehensive value score for the tag "participating annuity preference customer" is 50 points. During periods of high demand and limited resources at the beginning of the sales season, the second threshold is raised to 55 points in tandem with system load. If the score falls below 55, a judgment is triggered, and the tag is marked as a "cold tag." Subsequently, the system automatically suspends the scheduled feature engineering task for this tag and archives its results to a low-cost storage layer. Meanwhile, the first threshold remains at 75 points; both thresholds are uniformly generated and pushed by the same resource-aware threshold adjustment module.

[0046] As another concrete example, in the healthcare field, the comprehensive value score of the label "regular monitoring of mild hypertension" is 30 points. At this point, the overall system resource utilization is low, and the second threshold is at the baseline value of 35 points. When the score is below 35, the label is considered a "cold label." Its regular batch evaluation calculation task is then stopped, and historical evaluation results are migrated to cold data archive storage. The corresponding first threshold then remains at its baseline value of 55 points. Both threshold values ​​are driven by the same resource monitoring module and adaptively fluctuate with changes in platform load.

[0047] It is understandable that by employing a dual-threshold comparison and determination logic consisting of a first threshold and a second threshold in this embodiment, at least the following technical effects are achieved: First, the buffer between the two thresholds effectively avoids the frequent state fluctuations of tags in the single-threshold mode. Tags with scores between the first and second thresholds maintain their original state, and state transitions only occur when the value changes significantly enough and crosses the corresponding boundary, greatly improving the stability of tag hot and cold management.

[0048] Secondly, both the first and second thresholds are dynamically linked to the resource utilization rate of the tag management system, enabling the hot / cold tag judgment criteria to have adaptive and flexible adjustment capabilities. When system resources are scarce, both thresholds increase simultaneously, automatically excluding more low- and medium-value tags from the active computing system and prioritizing resource supply for high-value tags; when resources are abundant, the thresholds decrease, restoring a more lenient activity judgment standard, allowing more tags to respond to business needs in a timely manner. This judgment mechanism itself has an embedded closed-loop adjustment capability based on resource awareness, achieving dynamic self-balancing in tag lifecycle management without external manual configuration.

[0049] Step S140: When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. Specifically, in this step, after the tag's temperature status changes to "cold tag," the tag lifecycle manager will drive two types of operations. First, it sends a stop command to the task scheduling center, suspending all registered routine computation tasks for the tag, including offline batch processing tasks and near-line stream aggregation tasks, and unsubscribing these tasks from upstream data consumption in the dependency graph, immediately releasing computation slots. Second, it initiates a data migration process, moving the tag's generated result datasets (such as user lists, feature snapshots, and intermediate tables) and the tag's computational logic metadata from the current high-performance storage layer (such as SSD nodes in a distributed file system or an in-memory database) to low-cost storage media (such as an archive layer for erasure coding storage or a low-frequency access cloud object storage). The migration process ensures data integrity verification and records the archive path to the metadata center for addressing during subsequent wake-up calls.

[0050] As a specific example, in the life insurance field, after the label "potential customers of high-end health and wellness communities" becomes less popular, the system automatically stops its daily routine feature labeling task based on a distributed computing engine, and migrates the existing customer group snapshots and intermediate calculation results of this label from high-speed distributed storage to archived object storage.

[0051] As another concrete example, in the healthcare field, once the label of "high-risk diabetic foot patient" enters a cold state, its near real-time streaming computing tasks are suspended and no longer consume vital sign data streams; the generated patient risk scores and associated time-series monitoring data are transferred to an inexpensive cold data storage layer.

[0052] In some embodiments of the present invention, when a hot tag is determined, the routine calculation task is performed at a first update frequency and the tag result is stored in a first type of storage medium; Specifically, in this step, after the tag status is marked as "hot tag," the configuration management center can read the preset first update frequency parameter for this type of tag. This frequency can be set to a high frequency, such as once per hour, once every 15 minutes, or even near real-time streaming updates, to ensure that the tag results can quickly reflect changes in upstream data. The task scheduling engine generates a timed trigger plan for the tag's routine computation tasks based on this frequency and allocates sufficient computing resources. Simultaneously, the tag result data is stored on a first-type storage medium, characterized by high random read / write performance and low latency, typically such as a distributed file system based on solid-state drives, an in-memory database, or high-performance columnar storage, suitable for handling frequent reads and rapid updates. Data is written with a Time To Live (TTL) or version number for convenient hot data management.

[0053] As a concrete example, in the life insurance sector, the tag "potential customer for high-end senior living communities" is identified as a hot tag. The system sets the first update frequency for the daily customer characteristic tagging task for this tag to "incremental update once every 1 hour," and prioritizes its execution in the scheduling queue. The resulting customer segmentation results snapshot is directly written to the first type of storage medium, which can be a distributed in-memory database composed of all-flash nodes, allowing agents' sales apps and marketing platforms to query it in real time with sub-millisecond latency.

[0054] As another concrete example, in the healthcare field, the label "high-risk patient for diabetic foot" is identified as a hot label on a chronic disease management platform. Its risk score calculation task is configured with a first update frequency of "triggered micro-batch calculation every 5 minutes" to keep up with blood glucose data and electronic medical record updates in near real-time. The calculation results are stored on a first-class storage medium, which is a high-performance time-series in-memory database, ensuring a latency of less than 10 milliseconds when accessed by the clinic system.

[0055] When a label is identified as a cold label, the automatic archiving operation includes transferring the data associated with the preset label to a second type of storage medium with lower access performance than the first type of storage medium.

[0056] Specifically, in this step, the tag is identified as a "cold tag" and an automatic archiving operation is triggered. Besides suspending its routine computation tasks (already implemented in step S140), the data migration module migrates all datasets associated with the tag from its current hot storage layer to a second-type storage medium. The second-type storage medium is characterized by lower access performance than the first type, but significantly lower unit storage cost; typical examples include erasure coding-based archive file systems, low-frequency access or archive storage in cloud object storage, and tape libraries. During the migration, the system first scans the tag metadata to identify all relevant data tables, partitions, and indexes, generating a migration list; then, it performs data copying in an asynchronous batch processing manner while simultaneously verifying integrity; after the migration is complete, it updates the storage path pointer in the tag metadata to point to the new location in the second-type storage medium and cleans up the original hot storage layer copy, achieving storage degradation.

[0057] As a concrete example, in the life insurance sector, the tag "participating annuity preference customer" is considered a cold tag. The automatic archiving operation, besides suspending its original daily offline full-scale calculation task, also transfers its historical clustering results, intermediate feature wide tables, etc., from the first type of distributed storage layer based on solid-state drives to a second type of storage medium. This second type of storage medium can be an archive object storage bucket based on a regular hard disk drive with a high compression ratio. After the transfer is completed, this tag will no longer be loaded by default in the insurance marketing system reports corresponding to this tag, only retaining the ability to access it on demand.

[0058] As another concrete example, in the healthcare field, the tag "regular monitoring of mild hypertension" is identified as a cold tag. The system migrates its accumulated patient assessment results, tracking records, and intermediate statistical tables from a first-class high-performance columnar storage database to a second-class storage medium. This second-class storage medium can be a cold data archive storage layer, based on inexpensive cloud disks and limited to a maximum number of requests per second. After the migration, the data access entry point for this tag points to the archive layer, and it only receives a response through an unfreezing request when doctors actively query historical risk trajectories; the daily chronic disease management dashboard is no longer automatically loaded.

[0059] It is understandable that in this embodiment, by configuring differentiated update frequencies and storage media for tags based on their hot or cold status, the steps of this embodiment achieve refined hierarchical management of computing and storage resources. For hot tags, using a first update frequency and a first type of high-performance storage media ensures high timeliness and low-latency access to tag results, meeting the stringent requirements of core business scenarios for data freshness and response speed. For cold tags, during their automatic archiving operation, data is transferred from high-performance storage to a second type of storage media with lower access performance but significantly reduced costs, directly reducing space occupation and continuous energy consumption on expensive storage layers. This differentiated storage and computing strategy allows the tag system to concentrate high-value, high-activity tags while keeping total resource investment under control, achieving targeted cost reduction and efficiency improvement from both storage cost and update overhead dimensions, avoiding the waste caused by uniformly configuring high-performance resources for all tags.

[0060] Step S150: When a call request for the target cold tag is detected, an asynchronous wake-up process is initiated to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

[0061] Specifically, in this step, when the tag service provides a query or call interface, it can first check the status of the requested tag. If a cold tag is found, the synchronous request is intercepted and a "processing" status message is returned, while a wake-up event is sent to the asynchronous wake-up coordinator. The asynchronous wake-up process first reads the original calculation logic (such as algorithm container image and parameters) and the upstream data source information that the cold tag depends on from the archive metadata database. Then, if resources permit, a temporary calculation task is created as needed, and the calculation is re-executed using the latest upstream data to generate an updated tag result. The new result is written to the hot storage layer or cache, and the tag status is temporarily set to "wake-up". Subsequently, the wake-up coordinator intervenes in the value decay model, using this wake-up behavior as a positive incentive factor, directly adding a high-weight timeliness incentive value (such as directly adding 30 points to the original score or setting it above the protection line) to the comprehensive value score of the tag, ensuring that it will not be immediately archived again due to a low score within a subsequent preset protection window.

[0062] As a concrete example in the life insurance sector, when an agent requests to filter the list of "potential clients for high-end senior living communities" through a marketing platform, the system detects this inactive tag request and asynchronously loads its calculation logic. It then reconstructs the tag result by combining the latest asset and valuation data and injects it into the cache. Simultaneously, the agent's comprehensive value score is directly adjusted from 52 to 82, avoiding the need for re-archiving in the short term.

[0063] As another concrete example, in the healthcare field, when a doctor accesses a patient's risk label in the clinic system, the system identifies "high-risk patient for diabetic foot" as a cold label and immediately initiates asynchronous recalculation. It uses the latest physical examination and blood glucose data to score the patient, generates the latest risk classification, and displays it back. At the same time, it corrects the label's value score from 40 to 72 to ensure its stability and usability in the near future.

[0064] Understandably, the present invention, through the steps of "collecting behavioral indicator data from multiple dimensions associated with preset tags" and "applying time-decay weighting to the behavioral indicator data of each dimension based on a preset value decay model, and evaluating the comprehensive value score of the preset tag based on the weighted data," transforms the originally unquantifiable "tag value" into a calculable and comparable comprehensive value score. Time-decay weighting makes recent behavior contribute more significantly to the value, and the score dynamically reflects the tag's current activity level, rather than relying solely on static creation time or historical cumulative data. This step provides a quantitative basis for subsequent automated decision-making.

[0065] Furthermore, the present invention, through the step of "determining the temperature state of a preset label based on the comparison result of the comprehensive value score and the dynamic threshold," maps continuous value scores to discrete temperature states, establishing a decision logic of "high value corresponding to high resource guarantee, and low value corresponding to low resource consumption." The introduction of the dynamic threshold enables the judgment criteria to adaptively adjust with the overall resource pressure of the system, tightening the criteria when resources are scarce and relaxing the criteria when resources are abundant, ensuring that resource allocation always matches the actual carrying capacity of the system.

[0066] Furthermore, the present invention implements the temperature status decision into a substantial resource recovery action by stepping on the following: "When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered, including suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium." Suspending the calculation task directly releases the central processing unit and memory resources, while migrating the data to a low-cost storage medium releases expensive high-performance storage space, thereby achieving resource overhead control for low-value tags.

[0067] Furthermore, the present invention addresses the potential availability degradation issue following resource reclamation by initiating an asynchronous wake-up process upon detecting a call request for a target cold tag, thereby reconstructing the tag result of the target cold tag and correcting its overall value score. Asynchronous wake-up ensures that archived cold tags can still be reconstructed and have results returned when needed, rather than becoming completely unusable. Correcting the overall value score prevents woken-up tags from immediately triggering archiving again due to historically low scores, avoiding the jarring effect of wake-up immediately followed by archiving and providing tags with a reasonable period of active observation.

[0068] In summary, the present invention first quantifies the value of tags into comparable scores, and then automatically performs differentiated allocation and recycling of resources based on these scores. This further ensures that tags whose resources have been recycled can still be activated for service when needed. This closed loop allows the system to concentrate resources on a few high-value tags most of the time, while simultaneously recycling resources from a large number of low-value tags, significantly reducing overall resource overhead. Furthermore, when low-value tags are occasionally invoked by business applications, an asynchronous wake-up mechanism can be used to guarantee their availability, thus achieving the goal of "effectively controlling system resource overhead while ensuring tag availability."

[0069] In some embodiments of the present invention, the initiation of the asynchronous wake-up process to reconstruct the tag result of the target cold tag includes: Read the calculation logic of the target cold tag and the upstream data source information on which the target cold tag depends from the archive storage area of ​​the cold tag; Specifically, in this step, when the asynchronous wake-up process is triggered, the coordinator can be woken up to first initiate a query to the archive metadata service based on the unique identifier of the target cold tag. The archive storage area not only stores the historical result data of the tag, but also persistently saves the tag's computational logic definition (such as the algorithm package mirror address, feature engineering code library version, and main class parameters) and its dependent upstream data source information (including source database table names, partition fields, data stream topics, and filtering conditions). The reading process loads the computational logic code text and dependent metadata configuration into memory through the archive storage's access interface (such as reading from the archive layer of a distributed file system). After reading is complete, this metadata is parsed to construct an executable computational task description object, which is then passed to subsequent computational task triggers.

[0070] As a concrete example in the life insurance field, upon receiving an asynchronous wake-up request for the tag "potential customer for high-end senior living communities," the system reads the metadata file for that tag from its archived cold data bucket. This file contains a script containing the calculation logic. The core logic is to filter customers with assets greater than 5 million and who have browsed senior living communities in the past 6 months from the customer investable asset broad table. The upstream data sources include the "Customer Asset Summary Table (partitioned by day)" and the "Customer Behavior Track Log." The system reads this script and the source table information completely and parses out the required list of input tables and partitioning filtering strategies.

[0071] As another concrete example in the healthcare field, in response to a wake-up request tagged "high-risk diabetic foot patient," the system reads its computational logic from archived storage: the executable code package path and startup class of a computation job in the stream processing engine, as well as a feature computation configuration; upstream data sources include "real-time blood glucose monitoring data streams" and "glycated hemoglobin test record tables in the electronic medical record database." The system loads this metadata to confirm the latest availability status of the required data sources.

[0072] Based on the computational logic and the upstream data source information, computational tasks are triggered on demand to generate updated label results; Specifically, in this step, the task description object obtained from the task trigger can be computed, and a one-time computation job can be dynamically submitted to the computation resource scheduler. During execution, this job connects to the declared upstream data source, reads incremental data or a full snapshot within the latest time window, and strictly follows the original computation logic to generate labels. The computation framework can be an offline batch processing engine, a near-line micro-batch engine, or a stream processing engine, depending on the computation type originally defined for the labels. This computation is not a restoration of the original routine schedule, but rather an on-demand triggered task that releases computation resources upon completion. The generated label result is a dataset containing the target object identifier and its latest label value (such as customer ID and label category), and the result is temporarily stored locally or in intermediate storage on the execution node.

[0073] As a concrete example, in the life insurance sector, the scheduling engine submits a distributed offline computing job to the resource scheduling cluster based on the batch data processing scripts and upstream data source information it reads. The job takes the latest "Customer Asset Summary Table" partition of the day and the behavior logs of the past 6 months as input, executes filtering logic, and generates a list of customers who currently meet the criteria of "intended customers of high-end health and wellness communities" and their confidence scores. The calculation result contains approximately 50,000 records.

[0074] As another concrete example in the healthcare field, the system submits a standalone stream processing job to the stream processing engine cluster based on the executable package and configuration information. This job replays the most recent four hours of blood glucose monitoring data from the distributed data stream pipeline and performs rule matching against the latest glycated hemoglobin test result in the electronic medical record to recalculate the patient's risk level. The job ultimately produces an updated risk label result set, including patient ID, risk stratification, and score timestamp.

[0075] The updated tag results are written to the archive storage area or cache of the hot tags.

[0076] Specifically, in this step, after the computation task is successfully completed, the data writing module writes the updated tag result dataset generated in the previous step to the target storage location. This location is designated as either the "archive storage area for hot tags" or a cache. The "archive storage area for hot tags" refers to a high-performance online storage layer (such as a distributed in-memory database or an active partition in a solid-state drive storage cluster) where the hot tag result data resides, rather than cold archive storage; the "cache" refers to a distributed cache cluster. The write operation uses either overwrite or append mode, and sets the corresponding expiration time or version flag. After the write is complete, the tag service gateway's cache and routing table are updated. Subsequent calls for this tag will directly retrieve data from this hot storage layer or cache without triggering further computation.

[0077] As a concrete example in the life insurance sector, 50,000 results for "potential customers of high-end senior living communities" are batch-written into an in-memory database cluster that serves as a hot-tag storage area, overwriting the original old data shards for that tag. Simultaneously, the system refreshes the results to the query cache of the front-end sales application, enabling agents to receive the latest tag results within milliseconds for subsequent screening requests.

[0078] As another concrete example, in the healthcare field, updated risk label results for "high-risk patients with diabetic foot" are written to a high-speed key-value cache, using the patient ID as the key and the risk stratification as the value, with a TTL of 2 hours. When the clinic system queries again, it directly retrieves the latest results from this cache, ensuring real-time access for doctors. Simultaneously, a copy is asynchronously written to a relational storage table of hot tags for subsequent historical trend queries.

[0079] It is understandable that in this embodiment, by refining the asynchronous wake-up process into reading computational logic and dependencies from the archive area, triggering computation on demand, and finally writing to hot storage or cache, this embodiment achieves efficient and low-cost wake-up of cold tags. Firstly, persisting computational logic and upstream data source information in the archive storage area means that even if a tag is in a cold state, it does not need to retain any resident computational processes or high-frequency query records in the metadata server, maximizing the saving of system resources. Secondly, based on the on-demand triggered computational task mode, computational resources are only consumed for tag reconstruction when the business truly needs it, avoiding the continuous overhead of routine computations. Thirdly, writing the wake-up output directly to the hot tag storage area or high-speed cache keeps the initial call latency after a cold tag wake-up within an acceptable range, and subsequent repeated calls do not require rebuilding, significantly improving the availability and response speed of cold tags after wake-up. This mechanism achieves optimal utilization of storage and computational resources while ensuring that the business is always available.

[0080] In some embodiments of the present invention, the correction of the overall value score of the target cold label includes: The wake-up behavior is input as a positive incentive factor into the value decay model to improve the overall value score of the target cold tag, so that it is exempt from triggering the automatic archiving operation again within a preset period of time.

[0081] Specifically, in this step, a special rule can be preset within the value decay model: when the input event type is "wake-up behavior," this event does not participate in the regular time decay weighted calculation, but instead acts directly as a positive incentive factor on the current comprehensive value score. The specific processing logic is as follows: after completing the regular weighted decay calculation of each dimension and calculating the base score, the model checks whether there is a wake-up behavior event for this tag within the current evaluation period. If so, a preset incentive increment is added to the base score, or the base score is multiplied by an incentive coefficient greater than 1 to generate a corrected comprehensive value score. The value of this incentive increment or coefficient is calibrated to ensure that the corrected score is at least higher than the currently effective second threshold by a safety margin. This ensures that even if the score naturally decays over time during the subsequent preset period (determined by the protection duration in the model parameters, such as 24 hours, 72 hours, etc.), it will always remain above the second threshold and will not trigger automatic archiving again. The corrected score is directly overwritten into the tag's metadata storage as the basis for the next temperature status determination.

[0082] As a concrete example in the life insurance field, after the tag "potential customer for high-end senior living communities" is asynchronously reactivated, the value decay model detects that this tag has undergone a reactivation behavior within the current period. The base comprehensive value score of this tag, calculated using standard decay weighting, is 48 points, lower than the current second threshold of 55 points, and would normally be archived. The model uses the reactivation behavior as a positive incentive factor input, executing the "base score plus incentive increment" rule, with a preset incentive increment of 30 points, resulting in a corrected comprehensive value score of 78 points. 78 points is significantly higher than the second threshold of 55 points, and during the subsequent 72-hour natural decay process, this score will not fall below 55 points, thus ensuring that the tag is not automatically archived again within 72 hours. Agents can stably use this tag for customer screening during this window period without worrying about the tag suddenly becoming unusable.

[0083] As another concrete example in the healthcare field, when the label "high-risk patient with diabetic foot" is accessed and activated by a doctor, the model also detects the activation event. The base score for this label is 40 points, and the current second threshold is 50 points, already within the cold label judgment range. The model injects the activation behavior as a positive incentive factor, using a fixed incentive increment of 32 points, resulting in a corrected score of 72 points. 72 points is higher than the second threshold of 50 points. Within a preset 48-hour protection period, even if the label is not accessed again and the score continues to decline, it will remain above the second threshold, preventing it from being archived again. This ensures that doctors can continuously access patient risk labels during the current treatment and subsequent follow-up periods, ensuring uninterrupted clinical use.

[0084] Understandably, this embodiment improves the overall value score by inputting wake-up behavior as a positive incentive factor into the value decay model. This embodiment produces the following technical effects: the wake-up behavior is directly incorporated into the tag value assessment system, providing a one-time positive correction to the score as an incentive factor, allowing newly woken, cold tags to obtain a valid score higher than the archiving threshold. This mechanism naturally achieves the effect of preventing automatic archiving from being triggered again within a preset time period at the score level, without introducing additional exemption flags or hard-coded protection rules into the judgment logic, maintaining the self-consistency between value assessment and lifecycle management logic. Simultaneously, the correction only applies to the score value itself, and the length of the protection period is determined by both the incentive magnitude and the decay rate, allowing the tag to smoothly return to a normal decay trajectory after enjoying a reasonable activity window, avoiding oscillations in tag status management.

[0085] In some embodiments of the present invention, such as Figure 3 As shown, after the asynchronous wake-up process is completed, it also includes: Step S161: Continuously collect subsequent behavioral indicator data of the awakened tags and iteratively update the comprehensive value score of the awakened tags; Specifically, in this step, after the target cold tag completes result reconstruction through the asynchronous wake-up process, the tag is reintegrated into the regular behavior indicator monitoring system. The indicator collection module resumes collecting all-dimensional behavior events for the tag, including subsequent call frequency, business performance contribution, upstream data source update status, and user feedback behavior, with the collection method consistent with S110. Each time a preset evaluation cycle trigger time is reached (e.g., midnight every day or hourly), the value decay model automatically pulls the new behavior indicator events for the tag since the last evaluation, weights each event according to a preset time decay function, and iteratively calculates the result by combining it with the historical score of the previous cycle to generate an updated comprehensive value score. After the score is updated, it is written to the tag metadata storage for use by the temperature status determination service in the next determination cycle. This closed loop of continuous collection and iterative updates enables the value evaluation of awakened tags to dynamically reflect their actual usage after being awakened.

[0086] As a concrete example in the life insurance sector, after the cold tag "potential customers of high-end senior living communities" was activated by agents, the system continuously collected their subsequent behavioral indicators. Within the first week after activation, this tag was frequently used by multiple agents for customer screening and successfully drove targeted invitations to an online senior living community promotion event, resulting in a significant increase in attendance conversion rates. Simultaneously, the number of agents giving positive feedback on the tag's accuracy increased. When the daily evaluation cycle was triggered, the value decay model iteratively updated the tag's comprehensive value score based on these newly collected high-value behavioral events, with the score gradually climbing from a revised 78 points at the time of activation to 85 points.

[0087] As another concrete example, in the healthcare field, once the cold-labeled "high-risk diabetic foot patient" is accessed and activated by doctors in their clinic, the system continuously monitors their subsequent performance. During the follow-up period after activation, this label is repeatedly used by the chronic disease management platform to generate patient follow-up tasks, and multiple endocrinologists provide "accurate" effectiveness evaluations of its risk grading. In the hourly assessment tasks, the value decay model incorporates this new behavioral data into the iterative calculation, and the overall value score steadily rises from a revised 72 points to 80 points.

[0088] Step S162: When the comprehensive value score rises above the first threshold, the temperature status of the awakened tag is upgraded to a hot tag, and routine calculation tasks are resumed. Specifically, in this step, during each evaluation cycle, the temperature status determination service obtains the updated comprehensive value score of the awakened tag and compares it with the currently effective first threshold. When the comprehensive value score is detected to have risen from below the first threshold to above the first threshold, the determination service upgrades the temperature status field of the tag from its current state (which could be a warm tag or a cold tag still with a protection marker) to a "hot tag". After the status upgrade event is published via the message queue, the task scheduling engine receives the event and performs a recovery operation: re-registering the routine computing task for the tag from the archived task configuration, generating a scheduled plan according to the first update frequency corresponding to the hot tag, allocating computing resources, and starting task execution. Thus, the tag is completely removed from the archived state of the cold tag and returns to the complete online service system.

[0089] As a concrete example in the life insurance sector, the "potential customer for high-end senior living communities" tag, after being reactivated, saw its comprehensive value score climb to 85 points due to continuous high-frequency use and positive feedback. This occurred during a non-peak sales season, with ample system resources and the first threshold at the baseline of 60 points. With a score of 85 higher than 60, the system officially upgraded its status to a "hot tag." The scheduling engine then resumed the tag's original daily routine Spark feature labeling task, executing it according to the first update frequency of "incremental updates every hour," and the tag re-entered the full hot tag service state.

[0090] As another concrete example in the healthcare field, the "high-risk diabetic foot patient" label achieved a comprehensive value score of 80 points in the iterative assessment after being activated, while the first threshold dropped to 55 points due to the system load returning to normal. Since 80 points is higher than 55 points, the status upgrade logic is triggered, and the label is upgraded to a "hot label." The system then resumes its corresponding real-time stream computing task, reconnecting to the blood glucose monitoring data stream and electronic medical record update stream at a first update frequency of micro-batch processing every 5 minutes, continuously producing the latest risk assessment results.

[0091] Step S163: When the comprehensive value score decays to below the second threshold, the automatic archiving operation is retried.

[0092] Specifically, in this step, if a reactivated tag fails to demonstrate sustained activity in subsequent iterative evaluations (e.g., reduced call frequency, lack of new positive feedback, or prolonged stagnation of upstream data sources), its overall value score will gradually decline due to the decay function over time. When, in a given evaluation cycle, the updated overall value score falls below the currently effective second threshold, the temperature status determination service re-marks the tag as a "cold tag" and publishes a cold tag status change event. This event triggers the same automatic archiving process as in S140: upon receiving the event, the task scheduling center suspends the tag's current routine computation task, and the data migration module migrates the tag's associated result data and computation logic back to low-cost storage media. Thus, the tag completes its full lifecycle loop from "reactivation-activity-decay-re-archiving," and resources are reclaimed.

[0093] As a concrete example in the life insurance sector, the "participating annuity preference customer" tag, after its score was last activated and corrected, saw almost no new calls within a month due to a shift in business focus. Agents also provided no further feedback, and the upstream data source update frequency remained low. Over time, its comprehensive value score gradually declined from the corrected 68 points to 38 points, falling below the current second threshold of 40 points. The system determined it to be a "cold" tag, triggering automatic archiving again, suspending its calculation task, and migrating the data back to a low-cost archiving storage layer, releasing the occupied computing and storage resources.

[0094] As another concrete example in the healthcare field, after the "regular monitoring of mild hypertension" tag was activated, due to the completion of the related chronic disease management project, there were only sporadic calls in the following weeks, with zero doctor feedback and a longer update interval for upstream electronic medical records. The comprehensive value score decayed over time from a revised 60 points to 28 points, below the second threshold of 35 points. The system then retried the automatic archiving operation, stopped the routine batch evaluation task for this tag, and transferred its patient monitoring history data back to the cold data archiving layer, restoring the cold tag status.

[0095] Understandably, this embodiment forms a complete closed loop for the subsequent monitoring and management of awakened tags by continuously collecting subsequent behavioral indicators and iteratively updating their comprehensive value scores after the asynchronous wake-up process is completed. Firstly, the continuous collection and iterative update mechanism allows the tag's temperature status to dynamically respond to actual subsequent usage after being awakened, rather than relying on a one-time correction value at the time of wake-up, avoiding the crude management defects of tags being "permanently hot upon wake-up" or "one-time correction determining their entire lifespan." Secondly, when the score rises above the first threshold, it automatically upgrades to a hot tag and resumes routine calculation tasks, allowing tags that truly recover business value after being awakened to seamlessly return to the online service system, ensuring the timeliness and availability of high-value tags. Thirdly, when the score decays below the second threshold, automatic archiving is triggered again, ensuring that tags that were briefly awakened but lack sustained usage can be promptly reclaimed, preventing valueless tags from continuously occupying computing and storage resources. Overall, this embodiment endows tag lifecycle management with adaptive cyclical capabilities, enabling tags to dynamically migrate between cold and hot states based on actual value, achieving continuous optimization of resource allocation.

[0096] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. Software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0097] In one embodiment, a label management device is provided, which corresponds one-to-one with the label management method described in the above embodiments. For example... Figure 4 As shown, the tag management device includes a data acquisition module 410, a value assessment module 420, a status determination module 430, a data processing module 440, and a score correction module 450. Detailed descriptions of each functional module are as follows: The data acquisition module 410 is used to collect behavioral indicator data from multiple dimensions associated with preset tags; The value assessment module 420 is used to perform time decay weighting on the behavioral indicator data of each dimension based on a preset value decay model, and to evaluate the comprehensive value score of the preset label based on the weighted data. The status determination module 430 is used to determine the temperature status of the preset label based on the comparison result between the comprehensive value score and the dynamic threshold. Data processing module 440 is used to trigger an automatic archiving operation when the temperature status is determined to be a cold tag. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. The score correction module 450 is used to initiate an asynchronous wake-up process when a call request for a target cold tag is detected, in order to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

[0098] In one embodiment, the state determination module 430 is specifically used for: When the comprehensive value score is higher than the first threshold, it is determined to be a hot tag; When the comprehensive value score is lower than the second threshold, it is determined to be a cold label; The values ​​of the first threshold and the second threshold are dynamically adjusted according to the resource utilization rate of the tag management system.

[0099] In one embodiment, the state determination module 430 is further configured to: When a tag is identified as a hot tag, the routine calculation task is performed at the first update frequency and the tag result is stored in the first type of storage medium; When a label is identified as a cold label, the automatic archiving operation includes transferring the data associated with the preset label to a second type of storage medium with lower access performance than the first type of storage medium.

[0100] In one embodiment, the fraction correction module 450 is specifically used for: Read the calculation logic of the target cold tag and the upstream data source information on which the target cold tag depends from the archive storage area of ​​the cold tag; Based on the computational logic and the upstream data source information, computational tasks are triggered on demand to generate updated label results; The updated tag results are written to the archive storage area or cache of the hot tags.

[0101] In one embodiment, the fraction correction module 450 is further configured to: The wake-up behavior is input as a positive incentive factor into the value decay model to improve the overall value score of the target cold tag, so that it is exempt from triggering the automatic archiving operation again within a preset period of time.

[0102] In one embodiment, the fraction correction module 450 is further configured to: Continuously collect subsequent behavioral indicator data of awakened tags and iteratively update the comprehensive value score of the awakened tags; When the comprehensive value score rises above the first threshold, the temperature status of the awakened tag is upgraded to a hot tag, and routine calculation tasks are resumed. When the overall value score decays to below the second threshold, the automatic archiving operation is retried.

[0103] In one embodiment, the multiple dimensions of behavioral indicator data include indicator data representing the frequency of tag calls, indicator data representing the contribution improvement generated by the tag in business activities, indicator data representing the update status of the upstream data source on which the tag depends, and indicator data representing the user's feedback behavior on the tag.

[0104] Understandably, the present invention, through the steps of "collecting behavioral indicator data from multiple dimensions associated with preset tags" and "applying time-decay weighting to the behavioral indicator data of each dimension based on a preset value decay model, and evaluating the comprehensive value score of the preset tag based on the weighted data," transforms the originally unquantifiable "tag value" into a calculable and comparable comprehensive value score. Time-decay weighting makes recent behavior contribute more significantly to the value, and the score dynamically reflects the tag's current activity level, rather than relying solely on static creation time or historical cumulative data. This step provides a quantitative basis for subsequent automated decision-making.

[0105] Furthermore, the present invention, through the step of "determining the temperature state of a preset label based on the comparison result of the comprehensive value score and the dynamic threshold," maps continuous value scores to discrete temperature states, establishing a decision logic of "high value corresponding to high resource guarantee, and low value corresponding to low resource consumption." The introduction of the dynamic threshold enables the judgment criteria to adaptively adjust with the overall resource pressure of the system, tightening the criteria when resources are scarce and relaxing the criteria when resources are abundant, ensuring that resource allocation always matches the actual carrying capacity of the system.

[0106] Furthermore, the present invention implements the temperature status decision into a substantial resource recovery action by stepping on the following: "When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered, including suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium." Suspending the calculation task directly releases CPU and memory resources, while migrating the data to a low-cost storage medium releases expensive high-performance storage space, thereby achieving resource overhead control for low-value tags.

[0107] Furthermore, the present invention addresses the potential availability degradation issue following resource reclamation by initiating an asynchronous wake-up process upon detecting a call request for a target cold tag, thereby reconstructing the tag result of the target cold tag and correcting its overall value score. Asynchronous wake-up ensures that archived cold tags can still be reconstructed and have results returned when needed, rather than becoming completely unusable. Correcting the overall value score prevents woken-up tags from immediately triggering archiving again due to historically low scores, avoiding the jarring effect of wake-up immediately followed by archiving and providing tags with a reasonable period of active observation.

[0108] In summary, the present invention first quantifies the value of tags into comparable scores, and then automatically performs differentiated allocation and recycling of resources based on these scores. This further ensures that tags whose resources have been recycled can still be activated for service when needed. This closed loop allows the system to concentrate resources on a few high-value tags most of the time, while simultaneously recycling resources from a large number of low-value tags, significantly reducing overall resource overhead. Furthermore, when low-value tags are occasionally invoked by business applications, an asynchronous wake-up mechanism can be used to guarantee their availability, thus achieving the goal of "effectively controlling system resource overhead while ensuring tag availability."

[0109] Based on the above tag management methods, such as Figure 5 As shown in the diagram, this embodiment of the invention also provides a structural schematic of an apparatus for a tag management method, the apparatus including a processor 51 and a memory 52 coupled to the processor 51. The memory 52 stores a computer program, which, when executed by the processor 51, causes the processor 51 to perform the steps of the tag management method in the above embodiment.

[0110] For further details regarding the implementation of the above technical solution by the processor 51 in the device for the above-described label management method steps, please refer to the description of the label management method provided in the above-described embodiments of the invention, which will not be repeated here.

[0111] The processor 51 can also be called a CPU (Central Processing Unit). The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 can also be a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor, or the processor 51 can be any conventional processor.

[0112] like Figure 6As shown in the diagram, this embodiment of the invention also provides a schematic diagram of a computer-readable storage medium, on which a readable computer program 61 is stored. The computer program 61 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in various embodiments of the invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks or optical disks, ROM (Read-Only Memory), RAM (Random Access Memory), or terminal devices such as computers, servers, mobile phones, and tablets.

[0113] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, apparatuses, or modules, and may be electrical, mechanical, or other forms.

[0114] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] Furthermore, the functional modules in the various embodiments of this invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one 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 be stored in a computer-readable storage medium.

[0116] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0117] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium, or a semiconductor medium (e.g., SSD (solid state disk)).

[0118] The technical solution provided by the present invention has been described in detail above. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

[0119] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0120] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A label management method, characterized in that, include: Collect behavioral indicator data from multiple dimensions associated with preset tags; Based on a preset value decay model, the behavioral indicator data of each dimension are weighted by time decay, and the comprehensive value score of the preset label is evaluated based on the weighted data. The temperature status of the preset label is determined based on the comparison result between the comprehensive value score and the dynamic threshold. When the temperature status is determined to be a cold tag, an automatic archiving operation is triggered. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. When a call request for a target cold tag is detected, an asynchronous wake-up process is initiated to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

2. The label management method according to claim 1, characterized in that, The step of determining the temperature status of the label based on the comparison result of the comprehensive value score and the dynamic threshold includes: When the comprehensive value score is higher than the first threshold, it is determined to be a hot tag; When the comprehensive value score is lower than the second threshold, it is determined to be a cold label; The values ​​of the first threshold and the second threshold are dynamically adjusted according to the resource utilization rate of the tag management system.

3. The label management method according to claim 2, characterized in that: When a tag is identified as a hot tag, the routine calculation task is performed at the first update frequency and the tag result is stored in the first type of storage medium; When a label is identified as a cold label, the automatic archiving operation includes transferring the data associated with the preset label to a second type of storage medium with lower access performance than the first type of storage medium.

4. The label management method according to claim 1, characterized in that, The process of initiating the asynchronous wake-up procedure to reconstruct the tag result of the target cold tag includes: Read the calculation logic of the target cold tag and the upstream data source information on which the target cold tag depends from the archive storage area of ​​the cold tag; Based on the computational logic and the upstream data source information, computational tasks are triggered on demand to generate updated label results; The updated tag results are written to the archive storage area or cache of the hot tags.

5. The label management method according to claim 1, characterized in that, The correction of the overall value score of the target cold label includes: The wake-up behavior is input as a positive incentive factor into the value decay model to improve the overall value score of the target cold tag, so that it is exempt from triggering the automatic archiving operation again within a preset period of time.

6. The label management method according to claim 1, characterized in that, After the asynchronous wake-up process is completed, it also includes: Continuously collect subsequent behavioral indicator data of awakened tags and iteratively update the comprehensive value score of the awakened tags; When the comprehensive value score rises above the first threshold, the temperature status of the awakened tag is upgraded to a hot tag, and routine calculation tasks are resumed. When the overall value score decays to below the second threshold, the automatic archiving operation is retried.

7. A label management device, characterized in that, include: The data acquisition module is used to collect behavioral indicator data from multiple dimensions associated with preset tags; The value assessment module is used to perform time decay weighting on the behavioral indicator data of each dimension based on a preset value decay model, and to evaluate the comprehensive value score of the preset label based on the weighted data. The status determination module is used to determine the temperature status of the preset label based on the comparison result between the comprehensive value score and the dynamic threshold. The data processing module is used to trigger an automatic archiving operation when the temperature status is determined to be a cold tag. The automatic archiving operation includes suspending the routine calculation task corresponding to the preset tag and migrating the data associated with the preset tag to a low-cost storage medium. The score correction module is used to initiate an asynchronous wake-up process when a call request for a target cold tag is detected, in order to reconstruct the tag result of the target cold tag and correct the comprehensive value score of the target cold tag.

8. The label management device according to claim 7, characterized in that, The status determination module is specifically used for: When the comprehensive value score is higher than the first threshold, it is determined to be a hot tag; When the comprehensive value score is lower than the second threshold, it is determined to be a cold label; The values ​​of the first threshold and the second threshold are dynamically adjusted according to the resource utilization rate of the tag management system.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the tag management method as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the tag management method as described in any one of claims 1 to 6.

Images