Memory data processing method based on intelligent agent and related device

Abstract

The invention provides a memory data processing method based on an intelligent agent and a related device, the memory layer of the intelligent agent comprises at least two memory layers, the data access speeds of the at least two memory layers are different, and the method comprises the following steps: when an access behavior of a marked user for target memory data is detected, the target memory data is processed; calculating a popularity influence parameter of the marked user, wherein the popularity influence parameter is used for representing the influence degree of the access behavior on the access popularity of the target memory data; calculating a popularity outbreak probability of the target memory data according to the popularity influence parameter, wherein the popularity outbreak probability refers to a probability that the access popularity of the target memory data is greater than a preset threshold after a first preset time; and adjusting a memory layer where the target memory data is located according to the popularity outbreak probability. In this way, the memory layer where the memory data of the intelligent agent is located can be dynamically adjusted, and the response speed of the question-answering system is increased.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a memory data processing method and related apparatus based on intelligent agents. Background Technology

[0002] An agent is an intelligent entity capable of autonomously perceiving its environment, making decisions, and executing tasks. Memory is a crucial component affecting an agent's understanding capabilities. With the continuous increase in data, and given the limitation of context length, processing the agent's memory data becomes key to balancing inference efficiency and computational resources.

[0003] With the continuous development of artificial intelligence technology, the financial and economic fields have begun to explore and use intelligent agents to handle complex tasks in the industry. However, the timeliness and scale of financial data are technical challenges for intelligent agents in terms of memory, and the response speed is still slow in the specific application of question-answering systems. Summary of the Invention

[0004] This application provides a method and related apparatus for processing memory data based on an intelligent agent, aiming to dynamically adjust the memory layer where the intelligent agent's memory data is located and improve the response speed of the question-answering system.

[0005] In a first aspect, embodiments of this application provide a memory data processing method based on an intelligent agent, wherein the memory layer of the intelligent agent includes at least two memory layers, and the data access speeds of the at least two memory layers are different from each other; the method includes: When a marked user's access behavior to the target memory data is detected, the popularity influence parameter of the marked user is calculated. The popularity influence parameter is used to characterize the degree of influence of the access behavior on the access popularity of the target memory data. The probability of a surge in popularity of the target memory data is calculated based on the popularity influence parameters. The probability of a surge in popularity refers to the probability that the access popularity of the target memory data is greater than a preset threshold after a first preset time. The memory layer containing the target memory data is adjusted based on the probability of the heat outbreak.

[0006] Secondly, embodiments of this application provide a memory data processing device based on an intelligent agent, wherein the memory layer of the intelligent agent includes at least two memory layers, and the data access speeds of the at least two memory layers are different from each other; the device includes: The first calculation unit is used to calculate the popularity influence parameter of the marked user when the access behavior of the marked user to the target memory data is detected. The popularity influence parameter is used to characterize the degree of influence of the access behavior on the access popularity of the target memory data. The second calculation unit is used to calculate the heat surge probability of the target memory data based on the heat influence parameter. The heat surge probability refers to the probability that the access heat of the target memory data is greater than a preset threshold after a first preset time. The memory data adjustment unit is used to adjust the memory layer where the target memory data is located according to the probability of the heat outbreak.

[0007] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and one or more programs, the one or more programs being stored in the memory and configured to be executed by the processor, the programs including instructions for performing steps in the method as described in the first aspect of this application.

[0008] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program or instructions stored thereon, wherein the computer program or instructions, when executed by a processor, implement the steps of the method described in the first aspect of this application.

[0009] As can be seen in this embodiment, when a marked user's access behavior to target memory data is detected, a popularity influence parameter of the marked user is calculated. This popularity influence parameter characterizes the degree of influence of the access behavior on the access popularity of the target memory data. Then, the popularity burst probability of the target memory data is calculated based on the popularity influence parameter. The popularity burst probability refers to the probability that the access popularity of the target memory data exceeds a preset threshold after a first preset time. Finally, the memory layer where the target memory data is located is adjusted based on the popularity burst probability. In this way, the probability of the memory data being accessed in large quantities is predicted by the degree of influence of the marked user's access behavior on the popularity of the target memory data. This allows for the adjustment of the memory layer where the target memory data is located in advance during the window period of the memory data's popularity burst, thereby improving the response speed of the question-answering system when the memory data is accessed by a large number of users in a burst. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a system architecture block diagram provided in an embodiment of this application; Figure 2 This is another system architecture block diagram provided in the embodiments of this application; Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application; Figure 4 This is a structural block diagram of another electronic device provided in an embodiment of this application; Figure 5 This is a flowchart illustrating a memory data processing method based on an intelligent agent, as provided in an embodiment of this application. Figure 6 This is a structural block diagram of a memory data processing device based on an intelligent agent provided in an embodiment of this application; Figure 7 This is a structural block diagram of another agent-based memory data processing device provided in the embodiments of this application; Figure 8 This is a schematic diagram of the structure of another electronic device provided in the embodiments of this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0013] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0014] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0015] In related technologies, intelligent agents utilize Large Language Models (LLMs) as their core computing engine, possessing the ability to autonomously perceive the environment, make decisions, and execute tasks, making them suitable for question-answering scenarios across various vertical domains. In the financial and economic field, the timeliness and scale of data pose technical challenges to the memory capabilities of intelligent agents. Typically, massive amounts of memory data are managed collaboratively using Long Short-Term Memory (LSTM). Real-time state-aware data, such as price fluctuations and instantaneous trading volumes in financial markets, are treated as short-term memory, while historical data, such as historical market events and customer behavior preferences, are considered long-term memory. STM is usually integrated into the model's context window, relying on attention mechanisms and memory caching for fast responses. Long-term memory depends on external storage systems, and its retrieval speed is limited by the performance of external storage. In other words, under normal circumstances, the response speed of intelligent agents to short-term memory is significantly faster than that to long-term memory. However, in financial markets, situations frequently arise where specific events or times trigger simultaneous access to the same long-term memory by a large number of users within a short period. In such cases, the question-answering system needs to search the long-term memory from external storage for each user's access, further impacting the system's response speed under high concurrency pressure.

[0016] To address the aforementioned issues, embodiments of this application provide a method and related apparatus for processing memory data based on intelligent agents.

[0017] The system architecture involved in the embodiments of this application is described below.

[0018] In one embodiment, such as Figure 1 As shown, system 10 includes a first electronic device 11 and a second electronic device 12, which are communicatively connected. The first electronic device 11 can be a client, and the second electronic device 12 can be a server. When a user performs an access action on target memory data using the first electronic device 11, the first electronic device 11 sends the access action to the second electronic device 12, which then executes the agent-based memory data processing method as described in the embodiments of this application.

[0019] In another embodiment, such as Figure 2 As shown, system 10 includes only a first electronic device 11. The user performs an access behavior for target memory data on the first electronic device 11, and the first electronic device 11 performs the agent-based memory data processing method as described in the embodiments of this application.

[0020] Further, please refer to Figure 3 , Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application. For example... Figure 3As shown, the electronic device 30 includes at least two storage units 31, and the data access speeds of the at least two storage units 31 are different. Specifically, each storage unit 31 is implemented as a memory layer of an intelligent agent deployed within the electronic device 30. The electronic device 30 can be either the first electronic device 11 or the second electronic device 12 described above.

[0021] In another embodiment, such as Figure 4 As shown, the electronic device 30 includes a first storage unit 32 and a second storage unit 33, wherein the data access speed of the first storage unit 32 is greater than that of the second storage unit 33. The first storage unit 32 can be implemented as a low-latency, high-speed storage medium, such as a cache or DRAM (Dynamic Random Access Memory), used to store memory data with access frequency exceeding a preset threshold. The second storage unit 33 can be implemented as a low-cost, low-speed storage medium, such as an SSD (Solid State Drive), HDD (Hard Disk Drive), or a distributed storage cluster combining SSDs and HDDs, used to store memory data with access frequency below a preset threshold. The electronic device 30 can be either the first electronic device 11 or the second electronic device 12 described above.

[0022] The following describes a memory data processing method based on an intelligent agent, provided by an embodiment of this application.

[0023] Please see Figure 5 , Figure 5 This is a flowchart illustrating a memory data processing method based on an intelligent agent, as provided in an embodiment of this application. Figure 5 As shown, the method includes: S501, when a marked user's access behavior to the target memory data is detected, the popularity influence parameter of the marked user is calculated.

[0024] The term "marked user" refers to a high-value user identified through system authentication and filtering based on historical access records. Specifically, users can submit authentication applications for high-value users in the front-end application of the question-and-answer system. The system retrieves the user's access records for various types of memory data over a past period and counts the number of times the access popularity of a particular type of memory data exceeds a preset threshold within a first preset time period after the user accesses it. When this number exceeds the preset threshold, it indicates that the user's access behavior can, to some extent, drive or guide a surge in public access to the data. For example, the access behavior of professional analysts or investors can drive a surge in general user access. In this case, the user is marked as a high-value user, and a correspondence between user identity code, data type, and number of popularity surges is established and stored in the high-value user database. The data type refers to the classification of the memory data's domain or attribute. For example, if the memory data is "net profit," then its corresponding data type is "financial indicator." In practical applications, when a user's access behavior for a certain memory data is detected, the user is identified as a marked user based on their user identity code. If the data type of the memory data belongs to the target data type corresponding to the marked user, then the access behavior is determined to be the marked user's access behavior for the target memory data, triggering the calculation of the marked user's popularity influence parameters. The popularity impact parameter characterizes the degree of influence of the access behavior on the access popularity of the target memory data. Access popularity is a quantitative indicator of the effective attention paid to the target memory data by the general public within a first preset time period. The first preset time period can be a fixed value, such as 3 hours, or it can be flexibly adjusted based on the data type of the target memory data; no single limitation is imposed here.

[0025] Specifically, the access behavior can be a questioning behavior targeting a target question, and the target memory data is the data required by the agent to answer the target question.

[0026] S502, calculate the heat burst probability of the target memory data based on the heat influence parameter.

[0027] The probability of a heatwave outbreak refers to the probability that the access popularity of the target memory data exceeds a preset threshold after a first preset time. Specifically, the probability of a heatwave outbreak can be obtained by nonlinearly mapping the heatwave influence parameter using a preset function. The preset function can be a sigmoid function, which maps the heatwave influence parameter to the interval [0, 1], enhancing the distinguishability of intermediate intensity signals while avoiding excessive influence from extreme signals, thus obtaining the probability value.

[0028] S503, adjust the memory layer where the target memory data is located according to the probability of the heat outbreak.

[0029] Specifically, a target memory layer is determined based on the probability of a trending topic breakout and the probability interval corresponding to each memory layer. When the current memory layer and the target memory layer are different, the target memory data is adjusted to the target memory layer. Specifically, the memory layers include a first memory layer and a second memory layer. The data access speed of the first memory layer is greater than that of the second memory layer. When the probability of a trending topic breakout is greater than or equal to a preset probability, the target memory data is adjusted to the first memory layer; when the probability of a trending topic breakout is less than the preset probability, the target memory data is adjusted to the second memory layer.

[0030] As can be seen, in this embodiment, when a marked user's access behavior to target memory data is detected, a popularity influence parameter of the marked user is calculated. This popularity influence parameter characterizes the degree of influence of the access behavior on the access popularity of the target memory data. Then, the popularity burst probability of the target memory data is calculated based on the popularity influence parameter. The popularity burst probability refers to the probability that the access popularity of the target memory data exceeds a preset threshold after a first preset time. Finally, the memory layer where the target memory data is located is adjusted based on the popularity burst probability. In this way, the probability of the memory data being accessed in large quantities is predicted by the degree of influence of the marked user's access behavior on the popularity of the target memory data. This allows for the adjustment of the memory layer where the target memory data is located in advance during the window period of the memory data's popularity burst, thereby improving the response speed of the question-answering system when the memory data is accessed by a large number of users in a burst.

[0031] In one possible example, calculating the popularity influence parameter of the marked user includes: obtaining the target data type of the target memory data; determining the scenario adaptation parameter of the access behavior based on the target data type, wherein the scenario adaptation parameter is used to characterize the degree of adaptation of the access behavior to a typical scenario, wherein the typical scenario refers to a scenario in which the access popularity of the memory data of the target data type in the historical record is greater than a preset threshold; obtaining at least one operation behavior of the marked user when accessing the target memory data; calculating the operation depth parameter of the marked user based on the target data type and the at least one operation behavior, wherein the operation depth parameter is used to characterize the exploration depth of the marked user for the target memory data; and calculating the popularity influence parameter of the marked user based on the scenario adaptation parameter and the operation depth parameter.

[0032] In the financial and economic fields, explosive data access typically occurs in specific scenarios. For example, historical data records show that a company's financial indicator data experiences explosive access during the release of its financial report or when a major event occurs. Typical scenarios for this type of data include the release of a company's financial report or a major event. Based on this, by analyzing the degree of fit between access behavior and typical scenarios corresponding to the target data type, scenario fit parameters for this access behavior are calculated.

[0033] Furthermore, in the financial and economic field, professional analysts or investors, as labeled users, have a more acute perception of financial events or data. They are typically able to detect market signals and access target memory data before ordinary users. Based on this, the depth of exploration of target memory data can be analyzed by examining the operational behavior of labeled users when accessing it. This analysis yields operational depth parameters, indirectly reflecting the potential value and likelihood of explosive growth of the memory data. The operational behavior refers to the actions performed by users related to the memory data when accessing it, such as browsing behavior and duration, downloading, and annotation.

[0034] Furthermore, the scene adaptation weight and operation depth weight that have been pre-iterated are obtained, and the scene adaptation parameters and operation depth parameters are weighted and summed to obtain the popularity influence parameters of the marked users.

[0035] As can be seen in this example, the target data type of the target memory data is obtained; the scenario adaptation parameters for access behavior are determined based on the target data type; at least one operation behavior of the marked user when accessing the target memory data is obtained; the operation depth parameters of the marked user are determined based on the target data type and at least one operation behavior; and finally, the popularity impact parameters of the marked user are calculated based on the scenario adaptation parameters and the operation depth parameters. In this way, key features of the marked user's access behavior are captured from the two core dimensions of scenario adaptation and operation depth, providing reliable input for subsequent calculations of the probability of popularity surges, thereby supporting dynamic adjustment decisions for the memory data.

[0036] In one possible example, determining the scenario adaptation parameters of the access behavior based on the target data type includes: obtaining scenario features and probability weights of at least one typical scenario corresponding to the target data type based on a preset scenario information mapping relationship, wherein the scenario features include time window features and market event features, and the probability weights are used to characterize the probability that the access popularity of the memory data of the target data type in the typical scenario is greater than a preset threshold; obtaining scenario information corresponding to the access behavior, wherein the scenario information includes access time and market event records; matching the scenario information corresponding to the access behavior with the scenario features of each of the at least one typical scenario to determine the target typical scenario with the highest matching degree; and determining the probability weight of the target typical scenario as the scenario adaptation parameter of the access behavior.

[0037] The scenario information mapping relationship includes the mapping relationship between data types and typical scenario features and probability weights. The scenario features include time window features and market event features. The time window feature refers to the typical time when the memory data of this data type experiences a surge in popularity, such as "30 minutes before the release of the financial report to 1 hour after the release". The market event feature refers to the market event that occurs when the memory data of this data type experiences a surge in popularity, such as "the release of favorable industry policies".

[0038] The probability weight is used to characterize the likelihood that the access popularity of the target data type's memory data in the typical scenario exceeds a preset threshold. For example, taking the memory data "net profit" of the financial indicator category as an example, assume that its corresponding typical scenarios include Scenario 1 and Scenario 2. Scenario 1 has the time window feature of "30 minutes before the financial report is released to 1 hour after the release" and the market event feature of "the release of positive news." Scenario 2 has the time window feature of "within the last 3 working days of the quarter" and the market event feature of "no specific event." It can be seen that the probability of the memory data "net profit" experiencing a surge in popularity is higher in Scenario 1 than in Scenario 2; for example, it is 0.95 in Scenario 1 and 0.7 in Scenario 2. Further, the scenario information corresponding to the marked user access behavior is obtained, including access time and market event records, and then matched one by one with the scenario features of each typical scenario to calculate the matching degree. Specifically, the time window matching degree is determined based on the difference between the user's access time and the time window range corresponding to the typical scenario, and the market event matching degree is determined based on the difference between the access time and the time of occurrence of the market event associated with the target memory data. Finally, the matching degree between the access behavior and the typical scenario is calculated by the weight ratio of the time window feature and the market time feature. Further, after calculating the matching degree between the access behavior and all typical scenarios, the typical scenario with the highest matching degree is determined as the target typical scenario, for example, scenario one. Then, the probability weight of scenario one is determined as the scenario adaptation coefficient of the access behavior, i.e., 0.95.

[0039] As can be seen in this example, the system can accurately match the access behavior of marked users with typical scenarios with high historical popularity, calculate scenario adaptation parameters to reflect the potential for the access behavior to trigger subsequent popularity outbreaks, provide key scenario dimension support for the calculation of subsequent popularity impact parameters, and improve the accuracy of popularity outbreak prediction.

[0040] In one possible example, calculating the operation depth parameter of the marked user based on the target data type and the at least one operation behavior includes: determining the operation weight of each of the at least one operation behavior based on the target data type, the operation weight being used to characterize the degree of contribution of the operation behavior to the exploration depth of the memory data of the target data type; calculating the operation score of the marked user when accessing the target memory data based on the operation weight of each operation behavior; and calculating the operation depth parameter of the marked user based on the operation score.

[0041] The same operation on memory data of different data types has different operation weights. For example, the download of memory data of financial indicators and the download of memory data of single-value market indices have different operation weights. The former type of memory data, such as a company's quarterly financial report, usually contains multi-dimensional information such as assets and profits. After downloading, it usually needs to be further analyzed and contributes more to the depth of exploration. The latter type of memory data, such as the Shanghai Stock Exchange Composite Index, is a single-dimensional value. After downloading, it usually only needs to be simply statistically analyzed or plotted and contributes less to the depth of exploration.

[0042] Furthermore, different operations on the same data type have different weights. For example, browsing and downloading memorized financial indicator data have different weights; the former contributes less to the exploration depth and therefore has a lower weight, while the latter contributes more and has a higher weight. In addition, different levels of the same operation on the same data type also have different weights. For instance, browsing financial indicator data includes browsing duration; browsing durations shorter than a preset duration have a lower weight, while browsing durations longer than a preset duration have a higher weight.

[0043] The operation score is the sum of the operation weights of all user actions during a single access to the target memory data, used to comprehensively measure the total exploration depth of that access. Further, the maximum operation weight corresponding to the target data type is calculated, i.e., the highest weight of a single operation action under that type or the highest score of all possible operation combinations. Then, the ratio of the operation score to the maximum operation weight is calculated to obtain the operation depth parameter.

[0044] As can be seen in this example, the system calculates the operation depth parameter from two dimensions: the value difference of the operation behavior and the characteristic difference of the data type. This parameter reflects the depth of the user's exploration of the target memory data, providing key behavioral dimension support for the subsequent calculation of the heat influence parameter and improving the accuracy of heat outbreak prediction.

[0045] In one possible example, determining the operation weight of each operation behavior among the at least one operation behavior based on the target data type includes: obtaining operation behavior information corresponding to the target data type based on a preset operation behavior information mapping relationship; the operation behavior information includes multiple reference operation behaviors, an operation type corresponding to each reference operation behavior, and a weighting rule; the operation type includes basic operations and deep operations; the basic operations are used to characterize the marked user's weak research intent on the target memory data, and the deep operations are used to characterize the marked user's strong research intent on the target memory data; the weighting rule is used to limit the maximum value of the operation weight of the basic operations and the minimum value of the operation weight of the deep operations; adjusting the weighting rule according to the complexity level of the target data type; identifying the basic operations and deep operations among the at least one operation behavior based on the operation behavior information; and determining the operation weight of each operation behavior among the at least one operation behavior based on the adjusted weighting rule.

[0046] The operation behavior mapping relationship includes a mapping relationship between data types and operation behavior information. The operation behavior information includes multiple reference operation behaviors corresponding to the data type, the operation type corresponding to each reference operation behavior, and weighting rules. The operation types include basic operations and advanced operations. For example, if the target data type is a financial indicator type, its multiple reference operation behaviors include browsing, downloading, and annotation. The operation type corresponding to browsing is a basic operation, while the operation types corresponding to downloading and annotation are advanced operations.

[0047] The weighting rules are used to limit the maximum value of the weights of basic operations and the minimum value of the weights of deep operations. For example, the maximum weight of basic operations is limited to 0.2, and the minimum weight of deep operations is limited to 0.4. These rules can be flexibly adjusted based on the complexity level of the target data type. Specifically, for more complex data types, such as financial indicators, since the research value of basic operations on complex data is lower than that of deep operations, the maximum value of the weights of basic operations can be reduced, while the minimum value of the weights of deep operations can be increased, further strengthening the underlying logic of deep operations driving market activity. For simpler data types, such as single-index market indices, since the research value of operational behavior on simple data is limited, the maximum value of the weights of basic operations can be increased, while the minimum value of the weights of deep operations can be decreased, avoiding excessive differentiation of operational intentions that could lead to weight distortion.

[0048] In practical applications, basic and deep operations are identified based on user behavior when accessing target memory data. Combined with adjusted weighting rules, the operation weight for each operation is determined. For example, assuming the maximum weight of the basic operation before adjustment is 0.2, the minimum weight of the deep operation is 0.4, the original weight of browsing is 0.2, and the original weight of downloading is 0.4, due to the high complexity of the target memory data, the weighting rules need to be adjusted to allocate more reasonable operation weights. Assuming the maximum weight of the basic operation after adjustment is 0.15, and the minimum weight of the deep operation becomes 0.45, in some embodiments, based on the adjusted rule weights, the operation weight of browsing becomes 0.15, and the operation weight of downloading becomes 0.45, making the allocation of operation weights more reasonable and improving the reliability of the operation depth parameters.

[0049] As can be seen in this example, the system achieves precise linkage between data type complexity, operation type, and operation weight, ensuring that the operation weight can truly reflect the differences in the exploration depth of different types of memory data by the labeled user, providing key support for the calculation of subsequent operation depth parameters, and improving the accuracy of popularity burst prediction.

[0050] In one possible example, after adjusting the memory layer where the target memory data is located based on the probability of a heat outbreak, the method further includes: within a first preset time period, correcting the probability of a heat outbreak based on the access behavior of other marked users towards the target memory data; adjusting the memory layer where the target memory data is located based on the corrected probability of a heat outbreak; after the first preset time period, calculating the access heat of the target memory data; and adjusting the memory layer where the target memory data is located based on the access heat of the target memory data and the preset threshold.

[0051] The aforementioned probability of a heatwave only reflects the prediction signal of a single user's access behavior. Within the first preset time after the user's access behavior triggers the heatwave prediction, the probability of a heatwave can be corrected by the access behavior of other marked users, thereby further improving the prediction accuracy.

[0052] Specifically, if at least a preset number of marked users exhibit similar access behaviors to the target memory data within a first preset time period, and if the operation depth parameter of other marked users when accessing the target memory data is greater than or equal to a preset value, then it indicates that the memory data has become a research focus of common interest to high-value users, and the probability of a surge in popularity needs to be further amplified. For example, P1 = P0 × (1 + 0.2 × n), where P1 refers to the corrected probability of a surge in popularity, P0 refers to the initial probability of a surge in popularity, and n refers to the number of marked users who meet the conditions, where n ≤ 5 to avoid over-amplification.

[0053] Through the aforementioned probability correction, the storage level of the memory data is dynamically adjusted within a first preset time period to ensure an accurate match between resources and the potential for a surge in popularity. For example, before the correction, the target memory data was determined to be in the second memory layer based on the probability of a surge in popularity for the marked users. After the probability correction based on the access behavior of the marked user group, the probability of a surge in popularity was increased, and the memory layer where the target memory data was located became the first memory layer. This triggered the migration and adjustment of the memory data, improving the access speed of subsequent users to the target memory data.

[0054] After the first preset time, the actual access behavior of general users is quantified to evaluate the real access popularity of the target memory data within the first preset time. Based on the access popularity and preset threshold, the memory layer where the target memory data is located is adjusted.

[0055] As can be seen, in this example, the probability of a surge in popularity is adjusted based on the access behavior of other marked users within the first preset time period to dynamically adjust the memory layer where the target memory data is located. After the first preset time period, the memory layer of the target memory data is adjusted based on the actual access popularity of the target memory data. This not only captures the potential for popularity in advance through the group behavior of marked users, but also verifies the effectiveness of the decision through the actual access popularity of the general public, significantly improving the system's resource utilization and user experience.

[0056] In one possible example, calculating the access popularity of the target memory data includes: obtaining all access records for the target memory data within a first preset time period; for each access record, calculating the access popularity contribution value of the access record based on the operation behavior in the access record; and summing the access popularity values ​​of all access records within the first preset time period to obtain the access popularity of the target memory data.

[0057] Specifically, the step of calculating the popularity contribution value of each access record based on the operation behavior in the access record includes: for each access record, based on a preset popularity contribution value mapping relationship, obtaining the basic popularity contribution value and time decay coefficient of the access record according to the operation behavior in the access record; and calculating the popularity contribution value of the access record based on the basic popularity contribution value and time decay coefficient.

[0058] Specifically, user actions can be categorized into basic effective actions, deeply effective actions, and ineffective actions. Basic effective actions reflect genuine user engagement, such as opening a page or browsing for more than 30 seconds. Deeply effective actions reflect deeper user needs, such as downloading files or annotating metrics. Ineffective actions exclude accidental clicks or other non-genuine needs, such as closing a page within 5 seconds of opening it. Furthermore, the popularity contribution value corresponding to basic effective actions is less than that corresponding to deeply effective actions, and the popularity contribution value corresponding to ineffective actions is zero. Based on the popularity contribution value corresponding to each action, the basic popularity contribution value for that access record is calculated by summing these values. Then, a time decay coefficient is introduced and multiplied by the basic popularity contribution value to obtain the overall popularity contribution value for that access record. The time decay coefficient is used to correct for interference from repeated visits by the same user within a short period of time. For example, the coefficient for the first visit by the same user within 10 minutes is 1.0, the coefficient for the second visit is 0.6, and the coefficient for the third and subsequent visits is 0.3. This avoids a single user from flooding the screen with visits and increasing the popularity. Finally, the popularity contribution values ​​of all access records within the first preset time period are summed to obtain the access popularity of the target memory data.

[0059] As can be seen, in this example, the access popularity of the target memory data is obtained by calculating the sum of the popularity contribution values ​​of all access records of the target memory data within the first preset time period. The memory layer where the target memory data is located is corrected based on the popularity burst probability, thereby realizing dynamic optimization of resources.

[0060] It should be noted that the question-and-answer system involved in this application embodiment will prompt users to read and confirm the relevant agreement upon first use, allowing them to choose whether to authorize the system to obtain and use their access records for related data processing and provide related optimization services. If the user agrees, the system will process the data based on the authorization; if the user refuses, the system's basic functions can be used normally, but some data optimization services that rely on personal access records will not be provided. It is understood that the access records related to data processing in this application embodiment all come from users who have agreed to the authorization.

[0061] For examples consistent with the above embodiments, please refer to... Figure 6 , Figure 6This is a structural block diagram of a memory data processing device based on an intelligent agent provided in an embodiment of this application. The memory data processing device 60 based on an intelligent agent includes: a first calculation unit 601, used to calculate the popularity influence parameter of the marked user when the access behavior of the marked user to the target memory data is detected, the popularity influence parameter being used to characterize the degree of influence of the access behavior on the access popularity of the target memory data; a second calculation unit 602, used to calculate the popularity burst probability of the target memory data according to the popularity influence parameter, the popularity burst probability being the probability that the access popularity of the target memory data is greater than a preset threshold after a first preset time; and a memory data adjustment unit 603, used to adjust the memory layer where the target memory data is located according to the popularity burst probability.

[0062] In one possible example, regarding the calculation of the popularity influence parameter of the marked user, the first calculation unit 601 is specifically configured to: obtain the target data type of the target memory data; determine the scenario adaptation parameter of the access behavior based on the target data type, the scenario adaptation parameter being used to characterize the degree of adaptation between the access behavior and a typical scenario, the typical scenario referring to a scenario in the historical record where the access popularity of the memory data of the target data type is greater than a preset threshold; obtain at least one operation behavior of the marked user when accessing the target memory data; calculate the operation depth parameter of the marked user based on the target data type and the at least one operation behavior, the operation depth parameter being used to characterize the exploration depth of the marked user towards the target memory data; and calculate the popularity influence parameter of the marked user based on the scenario adaptation parameter and the operation depth parameter.

[0063] In one possible example, regarding the determination of the scenario adaptation parameters for the access behavior based on the target data type, the first calculation unit 601 is specifically configured to: obtain scenario features and probability weights of at least one typical scenario corresponding to the target data type based on a preset scenario information mapping relationship, wherein the scenario features include time window features and market event features, and the probability weights are used to characterize the probability that the access popularity of the memory data of the target data type in the typical scenario is greater than a preset threshold; obtain scenario information corresponding to the access behavior, wherein the scenario information includes access time and market event records; match the scenario information corresponding to the access behavior with the scenario features of each of the at least one typical scenario to determine the target typical scenario with the highest matching degree; and determine the probability weight of the target typical scenario as the scenario adaptation parameter for the access behavior.

[0064] In one possible example, in calculating the operation depth parameter of the marked user based on the target data type and the at least one operation behavior, the first calculation unit 601 is specifically configured to: determine the operation weight of each of the at least one operation behavior based on the target data type, the operation weight being used to characterize the degree of contribution of the operation behavior to the exploration depth of the memory data of the target data type; calculate the operation score of the marked user when accessing the target memory data based on the operation weight of each operation behavior; and calculate the operation depth parameter of the marked user based on the operation score.

[0065] In one possible example, regarding the determination of the operation weight of each operation behavior among the at least one operation behavior based on the target data type, the first calculation unit 601 is specifically configured to: obtain operation behavior information corresponding to the target data type based on a preset operation behavior information mapping relationship, wherein the operation behavior information includes multiple reference operation behaviors, an operation type corresponding to each reference operation behavior, and a weighting rule, wherein the operation type includes basic operations and deep operations, wherein the basic operations are used to characterize the marked user's weak research intent on the target memory data, and the deep operations are used to characterize the marked user's strong research intent on the target memory data, and the weighting rule is used to limit the maximum value of the operation weight of the basic operations and the minimum value of the operation weight of the deep operations; adjust the weighting rule according to the complexity level of the target data type; identify the basic operations and deep operations among the at least one operation behavior based on the operation behavior information; and determine the operation weight of each operation behavior among the at least one operation behavior based on the adjusted weighting rule.

[0066] In one possible example, after adjusting the memory layer where the target memory data is located based on the heat burst probability, the agent-based memory data processing device 60 is further configured to: within a first preset time period, correct the heat burst probability based on the access behavior of other marked users for the target memory data; adjust the memory layer where the target memory data is located based on the corrected heat burst probability; after the first preset time period, calculate the access heat of the target memory data; and adjust the memory layer where the target memory data is located based on the access heat of the target memory data and the preset threshold.

[0067] In one possible example, regarding the calculation of the access popularity of the target memory data, the agent-based memory data processing device 60 is specifically configured to: acquire all access records of the target memory data within a first preset time period; for each access record, calculate the popularity contribution value of the access record based on the operation behavior in the access record; and sum the popularity contribution values ​​of all access records within the first preset time period to obtain the access popularity of the target memory data.

[0068] It is understood that since the method embodiments and the device embodiments are different presentations of the same technical concept, the content of the method embodiment section in this application should be adapted to the device embodiment section in a synchronous manner, and will not be repeated here.

[0069] When using integrated units, such as Figure 7 As shown, Figure 7 This is a structural block diagram of another agent-based memory data processing device provided in the embodiments of this application. Figure 7 The agent-based memory data processing device 60 includes a processing module 62 and a communication module 61. The processing module 62 controls and manages the operations of the agent-based memory data processing device, for example, executing the steps of the first computing unit 601, the second computing unit 602, and the memory data adjustment unit 603, and / or executing other processes of the technology described herein. The communication module 61 supports interaction between the agent-based memory data processing device and other devices. Figure 7 As shown, the agent-based memory data processing device may further include a storage module 63, which is used to store the program code and data of the agent-based memory data processing device.

[0070] All relevant content in each scenario involved in the above method embodiments can be referenced from the functional descriptions of the corresponding functional modules, and will not be repeated here. The above-mentioned agent-based memory data processing device 60 can all execute the above-mentioned... Figure 5 The method for processing memory data based on intelligent agents is shown.

[0071] Based on the description of the above method and device embodiments, please refer to... Figure 8 , Figure 8 This is a schematic diagram of the structure of another electronic device provided in the embodiments of this application. Figure 8 The illustrated electronic device includes a memory 801, a processor 802, a communication interface 803, and a bus 804. The memory 801, processor 802, and communication interface 803 are interconnected via the bus 804. Specifically, the electronic device may refer to the first electronic device 11 or the second electronic device 12 in the above embodiments.

[0072] The memory 801 may be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM).

[0073] The memory 801 can store programs. When the program stored in the memory 801 is executed by the processor 802, the processor 802 and the communication interface 803 are used to execute the various steps of the agent-based memory data processing method of the present application embodiments.

[0074] The processor 802 may be a general-purpose central processing unit (CPU), microprocessor, application specific integrated circuit (ASIC), graphics processing unit (GPU), or one or more integrated circuits, used to execute related programs to achieve the functions required by the units in the electronic device of this application embodiment, or to execute the agent-based memory data processing method of this application method embodiment.

[0075] The processor 802 can also be an integrated circuit chip with signal processing capabilities. In implementation, each step of the agent-based memory data processing method of this application can be completed by the integrated logic circuits in the hardware of the processor 802 or by software instructions. The processor 802 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory 801. The processor 802 reads the information in the memory 801 and, in conjunction with its hardware, performs the functions required by the units included in the electronic device of this application embodiment, or executes the agent-based memory data processing method of this application method embodiment.

[0076] The communication interface 803 uses transceiver devices, such as, but not limited to, transceivers, to enable communication between electronic devices and other devices or communication networks. For example, data can be acquired through the communication interface 803.

[0077] Bus 804 may include a pathway for transmitting information between various components of an electronic device (e.g., memory 801, processor 802, communication interface 803).

[0078] It should be noted that, although Figure 8 The illustrated electronic device only shows the memory 801, processor 802, and communication interface 803. However, those skilled in the art should understand that in specific implementations, the electronic device may also include other components necessary for normal operation. Furthermore, depending on specific needs, those skilled in the art should understand that the electronic device may also include hardware components to implement other additional functions. Moreover, those skilled in the art should understand that the electronic device may only include the components necessary for implementing the embodiments of this application, and may not necessarily include... Figure 8 All the devices shown.

[0079] This application also provides a computer storage medium storing a computer program / instructions thereon, which, when executed by a processor, implements some or all of the steps of any of the methods described in the above method embodiments.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for example, the division of units is merely a logical functional division, and there may be other division methods in actual implementation; for example, multiple units 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 through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0081] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0082] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can 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 can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a read-only memory, or random access memory, or a magnetic medium, such as a floppy disk, hard disk, magnetic tape, magnetic disk, or an optical medium, such as a digital universal optical disc, or a semiconductor medium, such as a solid-state drive.

[0083] The above description is merely a specific implementation of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application. Therefore, the protection scope of the embodiments of this application should be determined by the protection scope of the claims.

[0084] The device embodiments described above are merely illustrative. The units and modules described as separate components may or may not be physically separate. Furthermore, some or all of the units and modules can be selected to achieve the purpose of this embodiment, depending on actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0085] While this application discloses the above information, it is not limited thereto. Any person skilled in the art can easily conceive of variations or substitutions without departing from the spirit and scope of this application, and can make various alterations and modifications, including combinations of the different functions and implementation steps described above, as well as software and hardware implementation methods, all of which are within the protection scope of this application.

Claims

1. A method for processing memory data based on intelligent agents, characterized in that, The agent's memory layer includes at least two memory layers, and the data access speeds of the at least two memory layers are different. The method includes: When a marked user's access behavior to the target memory data is detected, the popularity influence parameter of the marked user is calculated. The popularity influence parameter is used to characterize the degree of influence of the access behavior on the access popularity of the target memory data. The probability of a surge in popularity of the target memory data is calculated based on the popularity influence parameters. The probability of a surge in popularity refers to the probability that the access popularity of the target memory data is greater than a preset threshold after a first preset time. The memory layer containing the target memory data is adjusted based on the probability of the heat outbreak.

2. The method according to claim 1, characterized in that, The calculation of the influence parameters on the popularity of the marked users includes: Obtain the target data type of the target memory data; The scenario adaptation parameters of the access behavior are determined based on the target data type. The scenario adaptation parameters are used to characterize the degree of adaptation between the access behavior and typical scenarios. The typical scenario refers to a scenario in which the access popularity of the memory data of the target data type in the historical record is greater than a preset threshold. Acquire at least one operational behavior of the marked user when accessing the target memory data; The operation depth parameter of the marked user is calculated based on the target data type and the at least one operation behavior. The operation depth parameter is used to characterize the exploration depth of the marked user for the target memory data. The popularity impact parameter of the marked user is calculated based on the scenario adaptation parameter and the operation depth parameter.

3. The method according to claim 2, characterized in that, The step of determining the scenario adaptation parameters for the access behavior based on the target data type includes: Based on a preset scenario information mapping relationship, the scenario features and probability weights of at least one typical scenario corresponding to the target data type are obtained. The scenario features include time window features and market event features. The probability weights are used to characterize the probability that the access popularity of the memory data of the target data type in the typical scenario is greater than a preset threshold. Obtain the scenario information corresponding to the access behavior, including the access time and market event records; The scenario information corresponding to the access behavior is matched one by one with the scenario features of each of the at least one typical scenario to determine the target typical scenario with the highest matching degree. The probability weights of the target typical scenario are determined as the scenario adaptation parameters of the access behavior.

4. The method according to claim 2, characterized in that, The step of calculating the operation depth parameter of the marked user based on the target data type and the at least one operation behavior includes: The operation weight of each of the at least one operation is determined according to the target data type, and the operation weight is used to characterize the degree of contribution of the operation to the exploration depth of the memory data of the target data type; Calculate the operation score of the marked user when accessing the target memory data based on the operation weight of each operation; The operation depth parameter of the marked user is calculated based on the operation score.

5. The method according to claim 4, characterized in that, Determining the operation weight of each of the at least one operation based on the target data type includes: Based on a preset mapping relationship of operation behavior information, operation behavior information corresponding to the target data type is obtained. The operation behavior information includes multiple reference operation behaviors, operation type corresponding to each reference operation behavior, and weighting rules. The operation type includes basic operations and deep operations. The basic operations are used to characterize the weak research intention of the marked user on the target memory data, and the deep operations are used to characterize the strong research intention of the marked user on the target memory data. The weighting rules are used to limit the maximum value of the operation weight of the basic operations and the minimum value of the operation weight of the deep operations. The weighting rules are adjusted according to the complexity level of the target data type; Based on the operational behavior information, identify the basic operation and deep operation in the at least one operational behavior; Based on the adjusted weighting rules, the operation weight of each operation in the at least one operation is determined.

6. The method according to any one of claims 1-5, characterized in that, After adjusting the memory layer where the target memory data is located according to the probability of the heat outbreak, the method further includes: Within the first preset time period, the probability of a surge in popularity is adjusted based on the access behavior of other marked users to the target memory data. The memory layer containing the target memory data is adjusted based on the corrected probability of heat outbreaks. After the first preset time, the access frequency of the target memory data is calculated; The memory layer containing the target memory data is adjusted based on the access frequency of the target memory data and the preset threshold.

7. The method according to claim 6, characterized in that, The calculation of the access frequency of the target memory data includes: Obtain all access records for the target memory data within the first preset time period; For each access record, calculate the popularity contribution value of the access record based on the operation behavior in the access record; The access popularity of the target memory data is obtained by summing the popularity contribution values ​​of all access records within the first preset time period.

8. A memory data processing device based on an intelligent agent, characterized in that, The agent's memory layer includes at least two memory layers, and the data access speeds of the at least two memory layers are different. The device includes: The first calculation unit is used to calculate the popularity influence parameter of the marked user when the access behavior of the marked user to the target memory data is detected. The popularity influence parameter is used to characterize the degree of influence of the access behavior on the access popularity of the target memory data. The second calculation unit is used to calculate the heat surge probability of the target memory data based on the heat influence parameter. The heat surge probability refers to the probability that the access heat of the target memory data is greater than a preset threshold after a first preset time. The memory data adjustment unit is used to adjust the memory layer where the target memory data is located according to the probability of the heat outbreak.

9. An electronic device, characterized in that, It includes a processor, a memory, and one or more programs, said one or more programs being stored in the memory and configured to be executed by the processor, said programs including instructions for performing the steps in the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-7.

Images