Cold and hot data migration method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a cold and hot data migration method and device, computer equipment, a readable storage medium and a program product, and relates to the technical field of software development. The method comprises the steps of obtaining customer information in response to a trigger operation for a front-end application; inputting the customer information into a pre-trained customer behavior prediction model to obtain a customer behavior prediction result; comparing the customer behavior prediction result with a cold and hot data knowledge graph to obtain associated data corresponding to the customer behavior prediction result and a predicted cold and hot data type of the associated data; generating a migration strategy for the associated data according to the actual cold and hot data type of the associated data and the predicted cold and hot data type; the migration strategy is used for migrating the associated data to a first memory where hot data is located or migrating the associated data to a second memory where cold data is located. By adopting the method, the identification accuracy of cold and hot data can be improved.

Description

Technical Field

[0001] This application relates to the field of software development technology, and in particular to a method, apparatus, computer equipment, readable storage medium, and program product for cold and hot data migration. Background Technology

[0002] Hot data refers to data that is frequently accessed, so it requires fast access speeds to improve system performance. These data pages are typically located in high-performance but relatively small-capacity memory. Cold data, on the other hand, is data that is not frequently accessed. This data is generally placed in large-capacity but relatively low-performance memory to take full advantage of its large capacity and low cost. Related technologies employ pre-defined hot and cold data judgment rules to identify hot and cold data. The "hot data" determined by the rules is stored on high-performance storage devices, while the "cold data" is stored on low-cost storage media, thus achieving rational utilization of storage resources.

[0003] However, the existing hot and cold data scheduling methods, which rely on preset strategies to determine hot and cold data, suffer from limitations such as insufficient integration and real-time performance of the identification rules. These limitations can easily lead to misjudgments of hot and cold data by the system. Therefore, the accuracy of hot and cold data identification is relatively low. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for cold and hot data migration that can improve the accuracy of cold and hot data identification, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for cold and hot data migration, including:

[0006] Retrieve customer information in response to trigger actions from the front-end application;

[0007] The customer information is input into a pre-trained customer behavior prediction model to obtain the customer behavior prediction result; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data.

[0008] The customer behavior prediction results are compared with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0009] Based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

[0010] In one embodiment, prior to obtaining customer information in response to a triggering operation on the front-end application, the method further includes:

[0011] Obtain the correlation between various data corresponding to the front-end application, the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data;

[0012] A hot and cold data knowledge graph is constructed based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data.

[0013] In one embodiment, constructing a hot and cold data knowledge graph based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data, includes:

[0014] The customer is identified as the central node, and various data from the front-end application are identified as multiple radiating nodes;

[0015] Based on the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data, the edges between the central node and its adjacent radiating nodes are constructed.

[0016] Based on the relationships between various data corresponding to the front-end application, edges are constructed between the adjacent radiating nodes and other radiating nodes in the direction away from the central node to obtain a hot and cold data knowledge graph.

[0017] In one embodiment, generating a migration strategy for the associated data based on the actual hot and cold data types of the associated data and the predicted hot and cold data types includes:

[0018] If the actual cold / hot data type of the associated data is the cold data and the predicted cold / hot data type is the hot data, then the migration strategy is determined to be to migrate the associated data to the first memory where the hot data is located.

[0019] If the actual hot / cold data type of the associated data is hot data and the predicted hot / cold data type is cold data, then the migration strategy is determined to be to migrate the associated data to the second memory where the cold data is located.

[0020] In one embodiment, the hot and cold data migration method further includes:

[0021] Collect real-time customer behavior and access data related to the real-time customer behavior;

[0022] The real-time customer behavior is compared with the customer behavior prediction result to obtain the first comparison result;

[0023] The access data is compared with the associated data to obtain a second comparison result;

[0024] Based on the first comparison result and the second comparison result, the pre-trained customer behavior prediction model and the hot and cold data knowledge graph are optimized.

[0025] In one embodiment, the optimization process for the pre-trained customer behavior prediction model and the hot and cold data knowledge graph based on the first comparison result and the second comparison result includes:

[0026] If the first comparison result is lower than the preset hit threshold, the pre-trained customer behavior prediction model is trained using the real-time customer behavior to obtain an optimized customer behavior prediction model.

[0027] If the second comparison result is lower than the preset hit threshold, the hot and cold data knowledge graph is corrected based on the customer's real-time behavior and the access data to obtain the corrected hot and cold data knowledge graph.

[0028] Secondly, this application also provides a cold / hot data migration device, comprising:

[0029] The information acquisition module is used to acquire customer information in response to trigger operations on the front-end application;

[0030] The behavior prediction module is used to input the customer information into a pre-trained customer behavior prediction model to obtain customer behavior prediction results; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data.

[0031] The type prediction module is used to compare the customer behavior prediction results with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, and the predicted hot and cold data types of the associated data.

[0032] The data migration module is used to generate a migration strategy for the associated data based on the actual hot and cold data types and the predicted hot and cold data types; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

[0033] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0034] Retrieve customer information in response to trigger actions from the front-end application;

[0035] The customer information is input into a pre-trained customer behavior prediction model to obtain the customer behavior prediction result; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data.

[0036] The customer behavior prediction results are compared with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0037] Based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0039] Retrieve customer information in response to trigger actions from the front-end application;

[0040] The customer information is input into a pre-trained customer behavior prediction model to obtain the customer behavior prediction result; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data.

[0041] The customer behavior prediction results are compared with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0042] Based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

[0043] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0044] Retrieve customer information in response to trigger actions from the front-end application;

[0045] The customer information is input into a pre-trained customer behavior prediction model to obtain the customer behavior prediction result; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data.

[0046] The customer behavior prediction results are compared with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0047] Based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

[0048] The aforementioned hot and cold data migration method, apparatus, computer equipment, computer-readable storage medium, and computer program product acquire customer information in response to trigger operations on a front-end application for subsequent identification of customer intent. The customer information is input into a pre-trained customer behavior prediction model to obtain customer behavior prediction results, thereby identifying the customer's intent and laying the groundwork for subsequent data migration of the data required to convey that intent. The pre-trained customer prediction model is trained on an initial neural network based on historical customer operation data. The customer behavior prediction results are compared with a hot and cold data knowledge graph to obtain the associated data corresponding to the prediction results, as well as the predicted hot and cold data types of the associated data. Further, based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data. This migration strategy is used to migrate the associated data to a first memory location containing hot data, or to a second memory location containing cold data. By using neural networks and a hot and cold data knowledge graph, customer intent is identified, and the hot and cold data types of the associated data are intelligently predicted based on the relationships between data in the hot and cold data knowledge graph, the relationships between data and customer behavior, and the hot and cold data types of the data, thereby improving the accuracy of hot and cold data identification. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a schematic diagram of the hot and cold data migration process in related technologies;

[0051] Figure 2 This is a flowchart illustrating a hot and cold data migration method in one embodiment;

[0052] Figure 3 This is a flowchart illustrating the steps involved in constructing a knowledge graph of hot and cold data in one embodiment.

[0053] Figure 4 This is a schematic diagram of the structure of a knowledge graph for hot and cold data in one embodiment;

[0054] Figure 5 This is a flowchart illustrating the hot and cold data migration method in another embodiment;

[0055] Figure 6 This is a schematic diagram of the model training / knowledge graph construction process in one embodiment;

[0056] Figure 7 This is a schematic diagram of a scenario where customer behavior triggers intelligent prediction, recommendation, and migration based on hot and cold data, as shown in one embodiment.

[0057] Figure 8 This is a schematic diagram of a customer behavior-triggered intelligent prediction, recommendation, and migration process based on hot and cold data in one embodiment.

[0058] Figure 9 This is a flowchart illustrating a method for optimizing a smart agent model based on hot and cold data in one embodiment.

[0059] Figure 10 This is a schematic diagram comparing the optimization of the hot and cold data knowledge graph in one embodiment.

[0060] Figure 11 This is a structural block diagram of a hot and cold data migration device in one embodiment;

[0061] Figure 12 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] As described in the background section, the hot and cold data scheduling methods of related technologies suffer from low accuracy in identifying hot and cold data. The inventors have discovered that the reason for this problem lies in... Figure 1The diagram illustrates the hot and cold data migration process for related technologies. Current methods often employ preset hot and cold data judgment rules to identify hot and cold data. "Hot data" determined by the rules is stored on high-performance storage devices such as DDR (Double Data Rate SDRAM), while "cold data" is stored on low-cost storage media such as "NVM" (Non-Volatile Memory) to achieve rational utilization of storage resources. For example, basic customer information is stored as "hot data" on high-performance storage devices like DDR; three months of customer account details are stored as "cold data" on low-cost storage media like "NVM". When a customer logs in, the front-end retrieves relevant data based on the customer's front-end operations. Upon login, the application initiates a customer information query, retrieving "basic customer information" from DDR. When the customer queries their account, the application initiates a billing details query, retrieving "three months of customer account details" from NVM. The back-end, based on preset rules such as data access frequency, triggers the migration of "three months of customer account details" to hot data. The use of preset strategies for hot and cold data identification suffers from several drawbacks. These include insufficient integration and real-time performance of the identification rules, limitations, and system misjudgments of hot and cold data. This leads to wasted storage resources, reduced system response performance, and decreased system processing efficiency. Specifically, the identification rules suffer from insufficient integration: they typically focus on basic metrics such as access frequency and timestamps, neglecting data relevance to business scenarios (e.g., upstream and downstream data in business processes) and semantic features (e.g., order processing vs. fault reporting). This lack of business relevance results in inaccurate hot and cold data identification. For example, in the above scenario, "account information" might be classified as hot data. Furthermore, the identification rules lack real-time performance, lagging behind changes in data status: data access patterns change in real time, but the identification rules have long cycles (e.g., daily / weekly updates), failing to capture minute-level hotspot fluctuations. For example, in the above scenario, it's impossible to predict customer account query operations and to migrate "account information" to DDR in advance. Inaccurate hot / cold data identification leads to access performance latency: When accessed data is stored as cold data in NVM, access performance latency occurs regardless of whether the front-end application first migrates the cold data back to high-speed storage before responding or directly accesses the NVM. In the scenario above, querying data directly from NVM results in lower response efficiency than storing it in DDR. Limitations of preset rules lead to soaring maintenance costs: As business scenarios become more complex, the number of preset rules grows exponentially, causing a surge in maintenance costs. For example, the data of interest related to different business scenarios differs; limitations of preset rules necessitate rule redesign, resulting in insufficient flexibility.

[0064] Based on the above reasons, this application provides a method for hot and cold data migration, which combines deep learning and knowledge graphs to analyze the relationships between data and predicts the hot and cold data types by predicting customer behavior, aiming to improve the accuracy of hot and cold data identification.

[0065] In one embodiment, such as Figure 2 As shown, a method for cold and hot data migration is provided. This embodiment illustrates the application of this method to a server migration system. It is understood that this method can also be applied to terminals, and to systems including terminals and servers, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0066] Step S202: In response to a trigger operation for the front-end application, obtain customer information.

[0067] Among them, the front-end application can be a user interface based on a browser or mobile device, which implements business functions through a GUI (Graphical User Interface), such as an e-commerce website or application.

[0068] The triggering operation can be an explicit or implicit behavioral event initiated by the user on the front end, such as clicking a button or a page timeout.

[0069] Customer information can include customer identity information and behavioral information, such as customer accounts and customer login behavior information for front-end applications.

[0070] Optionally, the migration system responds to a customer's triggered operation on the front-end application by obtaining customer information, including the customer's identity information and behavioral information corresponding to the triggered operation, such as logging into the front-end application or clicking on a component of the front-end application, which is used as a basis for subsequent identification of customer intent.

[0071] Step 204: Input customer information into the pre-trained customer behavior prediction model to obtain customer behavior prediction results.

[0072] The pre-trained customer prediction model is trained on an initial neural network based on historical customer operation data. This historical customer operation data includes page table data, access frequency, access time, data size, data type, and front-end application data requirements.

[0073] Among them, the customer behavior prediction result can be the prediction result of the customer's behavior towards the front-end application in the next moment, output by the model.

[0074] Optionally, the migration system inputs customer information into a pre-trained customer behavior prediction model to obtain the customer behavior prediction results output by the pre-trained customer behavior prediction model, thereby identifying the customer's intent and laying the groundwork for subsequent data migration of the data required to determine the customer's intent.

[0075] Step 206: Compare the customer behavior prediction results with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0076] Among them, hot and cold data knowledge graphs can be semantic networks that describe entities and their relationships in a structured form, supporting complex association queries and reasoning, revealing implicit relationships between data (such as the "user-transaction-device" relationship), and supporting intelligent analysis (such as hot data access pattern mining and cold data archiving strategy optimization).

[0077] The associated data can be the data from front-end applications that the predicted customer behavior will involve in the next moment, corresponding to the customer behavior prediction result. For example, if the customer behavior prediction result is access to the account information for this year, then the associated data would be the customer's billing information for this year. The data types of the associated data can be hot or cold, including both. Hot data is frequently accessed data, so it requires fast access speeds to improve system performance. These data pages are typically located in high-performance but relatively small-capacity memory (such as DDR). Cold data is infrequently accessed data, and it is generally placed in larger-capacity but relatively lower-performance memory (such as non-volatile memory, NVM) to fully utilize its advantages of large capacity and low cost.

[0078] Optionally, the migration system compares the customer behavior prediction results with the hot and cold data knowledge graph, queries the data associated with the customer behavior in the real-time hot and cold data graph, and the hot and cold data types of the associated data recorded in the knowledge graph, thereby obtaining the associated data corresponding to the customer behavior prediction results, and the predicted hot and cold data types of the associated data.

[0079] Step S208: Generate a migration strategy for the associated data based on the actual hot and cold data types and the predicted hot and cold data types.

[0080] The migration strategy is used to migrate associated data to the first memory location where hot data resides, or to the second memory location where cold data resides. The first memory can be Double Data Rate (DDR) synchronous dynamic random access memory, a storage space directly addressable by the CPU. DDR is characterized by its high data access speed, on the nanosecond level, and its price per GB is 5-10 times that of NVM. The second memory can be high-speed storage (NVM), a non-volatile memory that retains data even after power failure, acting as a "storage device" (distinct from the "temporary cache" role of RAM). Its speed is on the microsecond level or lower, with a single device capacity reaching several TB, and its unit cost is lower than DRAM (Dynamic Random Access Memory).

[0081] The actual hot / cold data type can be the hot / cold data type of the associated data at the current moment, which is related to the memory where the associated data is located at the current moment. For example, if the associated data is currently stored in the first memory, then the actual hot / cold data type of the associated data is hot data.

[0082] Optionally, the migration system compares the actual hot and cold data types of the associated data with the predicted hot and cold data types, and generates a migration strategy for the associated data based on the comparison results.

[0083] In the aforementioned hot and cold data migration method, customer information is obtained in response to trigger operations on the front-end application for subsequent identification of customer intent. This customer information is then input into a pre-trained customer behavior prediction model to obtain customer behavior prediction results, thereby identifying the customer's intent and laying the groundwork for subsequent data migration of the data required to convey that intent. The pre-trained customer prediction model is trained on an initial neural network based on historical customer operation data. The customer behavior prediction results are compared with a hot and cold data knowledge graph to obtain the associated data corresponding to the prediction results, as well as the predicted hot and cold data types of the associated data. Further, based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data. This migration strategy is used to migrate the associated data to a first memory location containing hot data, or to a second memory location containing cold data. By using neural networks and a hot and cold data knowledge graph, customer intent is identified, and the hot and cold data types of the associated data are intelligently predicted based on the relationships between data in the hot and cold data knowledge graph, the relationships between data and customer behavior, and the hot and cold data types of the data, thereby improving the accuracy of hot and cold data identification.

[0084] In one exemplary embodiment, such as Figure 3As shown, step S202, in response to a trigger operation for the front-end application and before obtaining customer information, further includes steps 302 to 304. Wherein:

[0085] Step 302: Obtain the relationships between various data corresponding to the front-end application, the relationships between customer behavior and various data of the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data.

[0086] Among them, the various data corresponding to the front-end application can be a variety of data involved in the front-end application, such as account information, customer billing information, product information, order information, logistics information, production and sales information, etc.

[0087] The correlation between customer behavior and the predicted hot / cold data types of application data includes whether the predicted hot / cold data type of the application data corresponding to the customer behavior is cold data or hot data. For example, if the customer behavior is browsing product information within seven days, and the corresponding application data is the browsed product information, then the predicted hot / cold data type of the product information is hot data. As another example, if the customer behavior is browsing product information a week ago, and the corresponding application data is the browsed product information, then the predicted hot / cold data type of the product information is cold data.

[0088] Optionally, the migration system acquires the relationships between various data corresponding to the front-end application, the relationships between customer behavior and various data of the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data. For example, the relationships between various data corresponding to the front-end application include inclusion relationships, correlation relationships, etc., such as order information including ordered product information and logistics information; the relationships between customer behavior and various data of the front-end application, such as customer behavior browsing product information, then browsing product information is associated with the front-end application data of product information.

[0089] Step 304: Construct a hot and cold data knowledge graph based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data.

[0090] Optionally, the migration system uses various types of customer and front-end application data as nodes, and uses the relationships between customer behavior, various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data in the front-end application, and the relationships between customer behavior and predicted hot and cold data types of application data as edges to construct a hot and cold data knowledge graph.

[0091] In this embodiment, a hot and cold data knowledge graph is constructed by leveraging the various relationships between customer behavior and different types of data from the front-end application. This allows for an intuitive and systematic understanding of the structure and connections between different data sets. Besides expressing various relationships, the graph also incorporates customer behavior and data hot / cold status information, enabling intuitive identification and management of data states. This optimizes storage resource scheduling, data migration, and retrieval strategies, improving system response speed and resource utilization efficiency. Furthermore, through these relationships, the system can predict future changes in hot and cold data based on customer behavior. For example, recent browsing behavior might be predicted as hot data, while historical behavior might be predicted as cold data. This can assist in data preloading and cache scheduling, thereby achieving intelligent hot and cold data management.

[0092] In an exemplary embodiment, step S304 constructs a hot and cold data knowledge graph based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data, including:

[0093] The customer is identified as the central node, and various data from the front-end application are identified as multiple radiating nodes. Based on the relationship between customer behavior and various data from the front-end application, as well as the relationship between customer behavior and the predicted hot and cold data types of application data, edges are constructed between the central node and its adjacent radiating nodes. Based on the relationship between various data corresponding to the front-end application, edges are constructed between adjacent radiating nodes and other radiating nodes in the direction away from the central node, thus obtaining a hot and cold data knowledge graph.

[0094] In this context, a central node can be a node in a knowledge graph that only has edges pointing to other nodes, but not edges pointing from other nodes to itself, such as a node located at the vertex in a tree diagram.

[0095] Among them, the radiation nodes can be other nodes in the hot and cold data knowledge graph besides the central node.

[0096] Optionally, such as Figure 4 As shown, a structural diagram of the hot and cold data knowledge graph is provided. The migration system identifies the customer as the central node, such as using the customer's identity information as the central node, and identifies various data from the front-end application as multiple radiating nodes, such as... Figure 4The system uses historical orders, recent orders, production and sales products, marketing activities, logistics information, production and sales product details, activity details, production and sales product examples, past account information, three months of account information, and historical production and sales product examples as radiating nodes. The migration system constructs edges between the central node and its adjacent radiating nodes based on the relationships between customer behavior and various data from the front-end application, as well as the relationships between customer behavior and predicted hot and cold data types in the application data. Figure 4 There is a correlation between customer behavior such as orders placed within the past year and historical orders. For example, if a customer's behavior is to query past property rights, the corresponding data is a production and sales instance, which is considered cold data. Furthermore, based on the correlations between various data points in the front-end application, the migration system constructs edges between adjacent radiating nodes and other radiating nodes in directions away from the central node, such as... Figure 4 In this context, historical orders and production / sales information are both contained within each other. Therefore, by constructing edges between historical orders and production / sales information, and between historical orders and logistics information, we can obtain a knowledge graph of hot and cold data.

[0097] Specifically, in the hot and cold data knowledge graph, the data set as hot data for front-end applications includes: real-time customer transaction data, including customer order information for the current year, order logistics information (such as "paid," "in activation," "delivery," "installation"), etc.; customer behavior data, including product / related product details viewed within the past seven days, marketing activities / related activity details (such as limited-time discounts, coupons), etc.; customer asset data, including "in-use" product instances under customer ownership; and customer account information: customer pending payment account information, customer account information for the current year, etc. The data set as cold data for front-end applications includes: historical customer transaction data, including historical order information from one year ago, historical order logistics information (such as "paid," "in activation," "delivery," "installation"), etc.; historical customer behavior data, including product / related product details viewed one week ago, marketing activities / related activity details (such as limited-time discounts, coupons), etc.; historical customer asset data, including "invalid" product instances under customer ownership; and historical customer account information, including customer account information from previous years, etc.

[0098] Understandably, customers with different account information can be associated with their own hot and cold data knowledge graphs, thereby enabling personalized hot and cold data customization for each customer, further improving the identification of hot and cold data for different customers, and thus achieving more accurate hot and cold data migration.

[0099] In this embodiment, relational reasoning and graph structure are used to effectively express the complex relationships of customer data, making it easier to identify hot and cold data. The structural design of central nodes and radiating nodes maintains the hierarchy and clarity of the graph, which is helpful for data management and migration optimization. In addition, the logic of relational reasoning and hot and cold data discrimination enhances the system's automation and intelligent processing capabilities.

[0100] In an exemplary embodiment, step S208 generates a migration strategy for the associated data based on the actual hot and cold data types and the predicted hot and cold data types, including:

[0101] If the actual cold / hot data type of the associated data is cold data, and the predicted cold / hot data type is hot data, the migration strategy is to migrate the associated data to the first memory where the hot data is located; if the actual cold / hot data type of the associated data is hot data, and the predicted cold / hot data type is cold data, the migration strategy is to migrate the associated data to the second memory where the cold data is located.

[0102] Optionally, if the migration system determines that the actual hot / cold data type of the associated data is cold data, and the predicted hot / cold data type is hot data, the migration strategy is to migrate the associated data to the first memory location where the hot data resides. If the migration system determines that the actual hot / cold data type of the associated data is hot data, and the predicted hot / cold data type is cold data, the migration strategy is to migrate the associated data to the second memory location where the cold data resides. It is understood that if the migration system determines that both the actual hot / cold data type and the predicted hot / cold data type of the associated data are hot data, or both are cold data, there is no need to generate a migration strategy, and the associated data is stored in the current memory. Furthermore, the migration system can obtain the actual hot / cold data type of the associated data by checking the memory location of the associated data or by querying the hot / cold data type mapping table of various data in the front-end application.

[0103] In this embodiment, the system can dynamically adjust the data storage location based on the comparison between predicted hot and cold data types and actual hot and cold data types, effectively improving data access efficiency or saving storage costs. Frequently accessed (predicted as hot data) cold data is placed in fast-access memory to improve access efficiency, while infrequently accessed (predicted as cold data) hot data is migrated to low-cost storage (cold memory) to save storage resources. This enables prediction-based automated hot and cold data migration, improving the system's intelligence level.

[0104] In one exemplary embodiment, the cold and hot data migration method described above further includes:

[0105] Collect real-time customer behavior and access data related to real-time customer behavior; compare the real-time customer behavior with the customer behavior prediction results to obtain the first comparison result; compare the access data with the associated data to obtain the second comparison result; and optimize the pre-trained customer behavior prediction model and the hot and cold data knowledge graph based on the first and second comparison results.

[0106] Among them, real-time customer behavior can be the real-time behavior taken by the customer in relation to the front-end application, and access data can be the data of the front-end application associated with the component clicked by the customer in real-time behavior, or the data of the front-end application displayed on the page, etc.

[0107] The first comparison result can be the similarity between real-time customer behavior and customer behavior; the second comparison result can be the similarity between access data and related data.

[0108] Optionally, each time a customer triggers an operation on the front-end application, the migration system collects the customer's real-time behavior and the access data involved in that behavior. This is equivalent to real-time tracking of the events triggered by the customer. The system compares the customer's real-time behavior with the predicted customer behavior to obtain a first comparison result, and compares the access data with related data to obtain a second comparison result. Furthermore, based on the first and second comparison results as feedback, the pre-trained customer behavior prediction model and the hot and cold data knowledge graph are optimized. It is understood that the content collected by the migration system also includes tracking subsequent application data access and response, such as subsequent data access events and related hot and cold migration events.

[0109] In this embodiment, when a customer triggers an action, customer behavior and related access data are collected and monitored in real time. The actual customer behavior is compared with a pre-trained behavior prediction model to obtain a first comparison result, evaluating the model's predictive performance. The actual access data is then compared with related data (such as data relationships in a knowledge graph) to obtain a second comparison result, evaluating the accuracy and matching degree of the relationships. Using the comparison results as feedback, the pre-trained model and knowledge graph are dynamically adjusted and optimized, thereby improving the accuracy of future behavior prediction and hot / cold data identification. Furthermore, in subsequent applications, data access and migration events are continuously monitored to verify the optimization effect and further adjust the strategy.

[0110] In an exemplary embodiment, the steps of the above embodiment, based on the first comparison result and the second comparison result, optimize the pre-trained customer behavior prediction model and the hot and cold data knowledge graph, including:

[0111] If the first comparison result is lower than the preset hit threshold, the pre-trained customer behavior prediction model is trained using real-time customer behavior to obtain an optimized customer behavior prediction model. If the second comparison result is lower than the preset hit threshold, the hot and cold data knowledge graph is corrected based on real-time customer behavior and access data to obtain a corrected hot and cold data knowledge graph.

[0112] The hit threshold can be the accuracy threshold set according to the prediction needs of the pre-trained customer behavior prediction model and the quantified cold and hot data knowledge graph.

[0113] Optionally, if the migration system determines that the first comparison result is lower than the preset hit threshold, real-time customer behavior is used as new training data to train the pre-trained customer behavior prediction model to obtain an optimized customer behavior prediction model; if the migration system determines that the second comparison result is lower than the preset hit threshold, the relationship between customer behavior and front-end application data in the hot and cold data knowledge graph, as well as the relationship between hot and cold data types, are corrected based on real-time customer behavior and access data to obtain a corrected hot and cold data knowledge graph.

[0114] In this embodiment, the prediction model is dynamically updated using the latest customer behavior to improve the accuracy and adaptability of predictions. Relationships in the knowledge graph are continuously corrected through real-time data to ensure the timeliness and accuracy of relationships. Combined with performance monitoring and self-optimization mechanisms, the intelligence level of data management is improved and human intervention is reduced. In addition, more accurate hot and cold data analysis, more reasonable storage strategies, and a better customer service experience are achieved.

[0115] In one exemplary embodiment, such as Figure 5 As shown, another method for cold and hot data migration is provided, including:

[0116] Intelligent model parameter collection is used to collect the parameters required for the deep learning model (pre-trained customer behavior prediction model): the initial parameters mainly include page table data, access frequency, access time, data size, data type, front-end application data requirements, and user behavior pre-judgment.

[0117] Specifically, technicians pre-build hot and cold data recommendation and transfer intelligence agents (pre-trained customer behavior prediction models) and hot and cold data knowledge graphs through model training / knowledge graph construction. For example... Figure 6 As shown, a flowchart illustrating the model training / knowledge graph construction process is provided.

[0118] Hot and cold data knowledge graphs are semantic networks that describe entities and their relationships in a structured form, supporting complex relational queries and reasoning.

[0119] The hot and cold data recommendation and migration intelligent agent is used for: For access data prediction and recommendation: Based on collected feature data parameters, it uses machine learning algorithms to predict the data that the front-end application may need to access, and predicts the hotness or coldness of this type of data; For hot and cold data migration: Based on the predicted recommendation results, it judges the matching degree with the existing hot and cold data distribution. If the recommendation does not match the current data distribution, it dynamically adjusts the migration strategy, generates hot and cold data migration instructions, and performs the hot and cold data migration to adapt to the ever-changing business scenario requirements; And For recommendation and migration result tracking: It tracks the predicted recommendation results and migration results, and submits the tracking results to the intelligent model training module.

[0120] Specifically, the front-end application initiates data access, memory management locates the data memory address, and data response feedback triggers hot and cold data recommendation and migration agent analysis. Relevant model parameters are collected, including page table data, access frequency, access time, data size, data type, front-end application data requirements, and user behavior prediction. Parameter knowledge is queried: based on the collected parameters, related knowledge is queried from the hot and cold data knowledge graph. Customer intent is identified based on relevant parameters and knowledge information. The data required by the application later is predicted based on the customer intent. The hot and cold data knowledge graph is used to identify the hotness or coldness of the predicted data. If the predicted data to be hot data for the front-end application is currently cold data, the agent automatically initiates a pre-warming data migration request. A pre-warming data migration instruction is generated and sent to heterogeneous memory management. The hot and cold data migration is completed, and the migration results are fed back.

[0121] The intelligent agent and knowledge graph are optimized by forming a closed loop of "recognition-transfer-feedback-optimization" through model training.

[0122] Specifically, after each "hot and cold data recommendation and migration" event processed by the agent, corresponding event processing performance tracking will be initiated. The "Recommendation and Migration Result Performance Tracking" module tracks subsequent application data access and response, such as subsequent data access events and related hot and cold data migration events. It analyzes the agent's predicted access data hit rate and the effectiveness of hot and cold data migration. The performance tracking records are submitted to the model training module. Combined with continuous algorithm optimization, parameter tuning, and hot and cold data knowledge graph revision based on the model training results.

[0123] In this embodiment, to address the data access needs of front-end applications, a hot and cold data recommendation and migration agent, along with a hot and cold data knowledge graph, is introduced. Through machine learning, deep learning models, and the hot and cold data knowledge graph, the system proactively analyzes the relationships between data and predicts the data required by the front-end application, identifying the predicted data's hotness or coldness. If the predicted data required by the application is cold data, a pre-warming data migration is automatically initiated (e.g., when a user browses a "mobile phone" product page, the agent predicts that the associated "accessories" data may become hot data and migrates it to high-speed storage in advance). This improves the response speed of front-end application data access and maintains system performance stability. This system, combined with the hot and cold data knowledge graph, uses intelligent model training to continuously optimize the hot and cold data recommendation and migration agent. It automatically adjusts feature weights based on historical migration results, forming a closed loop of "identification-migration-feedback-optimization," enabling the hot and cold data recommendation and migration agent to quickly adapt to new scenarios and reduce redundant development costs.

[0124] In one embodiment, such as Figure 7 As shown, and Figure 8 As shown, a scenario for intelligent prediction, recommendation, and migration based on customer behavior-triggered hot and cold data is provided, including:

[0125] Heterogeneous Memory Management: "Customer's three-month account details" are stored as "cold data" on low-cost storage media such as "NVM". Customer A habitually checks their monthly bill details before making online top-ups. When Customer A logs in, the agent predicts based on customer behavior that the customer may need to check their account details, and preemptively converts the "three-month account details" into "hot data" and migrates it to "DDR" (Memory Memory), thereby improving the front-end application's data access response speed and increasing access efficiency. When the customer logs in, relevant data is acquired based on the customer's front-end operations. These operations include: customer login, application initiating customer information query, and retrieving "customer basic information" from DDR; agent function: intent recognition and access data prediction; intent recognition: based on the customer's habit of checking monthly bill details before making online top-ups, predicting that the customer will need to check account details.

[0126] Predicted access data: "Customer's three-month account details"; The agent predicts the data and identifies its popularity: The agent identifies the popularity of the predicted data based on a knowledge graph of hot and cold data. The agent can consider multiple factors to judge the popularity of the data. If it finds that "Customer's three-month account details" is cold data, it initiates a pre-migration instruction; Agent: The hot and cold data migration instruction is generated and sent to the heterogeneous memory management; The heterogeneous memory management migrates "Customer's three-month account details" to DDR memory.

[0127] Front-end operation: When a customer inquires about their account, the application initiates a billing details query, retrieves the "customer's three-month account details" from the DDR, and recommends and migrates results. Performance tracking shows that the predicted data hit rate is 100%.

[0128] In one embodiment, such as Figure 9 As shown, a method for optimizing the hot and cold data agent model in the above embodiments is provided, including:

[0129] Heterogeneous memory management: Marketing campaign A is stored as "cold data" on low-cost storage media such as "NVM". Customer A recently browsed marketing campaign A, logged in, but the agent did not recognize the customer's intent. The customer then went to "NVM" to query marketing campaign A and its details. Subsequently, based on the performance tracking of recommendation and migration results, the agent and knowledge graph were optimized.

[0130] When a customer logs in, relevant data is obtained based on the customer's front-end operations: Front-end operations: When a customer logs in, the application initiates a customer information query and obtains "customer basic information" from DDR; Intelligent agent: From intent recognition to access data prediction; Intent recognition: Based on the customer's habit of checking bill details every month and then making an online recharge, predict that the customer will check account details.

[0131] Predicted access data: "Customer's three-month account details". Recommendation and migration result performance tracking: Front-end operations: When a customer performs a query for marketing campaign A, intent recognition fails, and the customer performs another query for marketing campaign A via NVM; Agent and knowledge graph optimization, including model training and tuning.

[0132] Optimize intent recognition algorithm: Add intent recognition based on the customer's recent marketing activity browsing history, specifically for the intent to "query marketing activity A". For example... Figure 10 As shown, a comparison chart of the hot and cold data knowledge graph before and after optimization is provided; optimization of the hot and cold data knowledge graph: the original "marketing campaign" is decomposed into marketing campaign (viewed in seven days) and historical marketing campaign (viewed one week ago).

[0133] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0134] Based on the same inventive concept, this application also provides a cold and hot data migration apparatus for implementing the aforementioned cold and hot data migration method. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more cold and hot data migration apparatus embodiments provided below can be found in the limitations of the cold and hot data migration method described above, and will not be repeated here.

[0135] In one exemplary embodiment, such as Figure 11 As shown, a hot and cold data migration device 1100 is provided, including: an information acquisition module 1101, a behavior prediction module 1102, a type prediction module 1103, and a data migration module 1104, wherein:

[0136] The information acquisition module 1101 is used to acquire customer information in response to a trigger operation on the front-end application;

[0137] The behavior prediction module 1102 is used to input customer information into a pre-trained customer behavior prediction model to obtain customer behavior prediction results; the pre-trained customer prediction model is obtained by training an initial neural network based on customer historical operation data.

[0138] The type prediction module 1103 is used to compare the customer behavior prediction results with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data.

[0139] The data migration module 1104 is used to generate a migration strategy for the associated data based on the actual hot and cold data types and the predicted hot and cold data types. The migration strategy is used to migrate the associated data to the first memory where the hot data is located, or to the second memory where the cold data is located.

[0140] Furthermore, in one embodiment, the hot and cold data migration device 1100 further includes a graph construction module, used to obtain the correlation between various data corresponding to the front-end application, the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data; and to construct a hot and cold data knowledge graph based on customer behavior, various data of the front-end application, the correlation between various data corresponding to the front-end application, the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data.

[0141] Furthermore, in one embodiment, the graph construction module is also used to identify customers as central nodes and various types of data from the front-end application as multiple radiating nodes; construct edges between the central node and its adjacent radiating nodes based on the correlation between customer behavior and various types of data from the front-end application, as well as the correlation between customer behavior and the predicted hot and cold data types of application data; and construct edges between adjacent radiating nodes and other radiating nodes in the direction away from the central node based on the correlation between various types of data corresponding to the front-end application, thereby obtaining a hot and cold data knowledge graph.

[0142] Furthermore, in one embodiment, the data migration module 1104 is further configured to determine a migration strategy of migrating the associated data to the first memory where the hot data is located when the actual cold and hot data type of the associated data is cold data and the predicted cold and hot data type is hot data; and to determine a migration strategy of migrating the associated data to the second memory where the cold data is located when the actual cold and hot data type of the associated data is hot data and the predicted cold and hot data type is cold data.

[0143] Furthermore, in one embodiment, the hot and cold data migration device 1100 further includes an intelligent optimization module, which is used to collect real-time customer behavior and access data related to real-time customer behavior; compare the real-time customer behavior with the customer behavior prediction result to obtain a first comparison result; compare the access data with the associated data to obtain a second comparison result; and optimize the pre-trained customer behavior prediction model and the hot and cold data knowledge graph based on the first comparison result and the second comparison result.

[0144] Furthermore, in one embodiment, the intelligent optimization module is also used to train the pre-trained customer behavior prediction model using real-time customer behavior when the first comparison result is lower than a preset hit threshold, to obtain an optimized customer behavior prediction model; and to correct the hot and cold data knowledge graph based on real-time customer behavior and access data when the second comparison result is lower than the preset hit threshold, to obtain a corrected hot and cold data knowledge graph.

[0145] Each module in the aforementioned hot and cold data migration device 1100 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0146] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 12As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores customer information, hot and cold data knowledge graphs, related data, and actual hot and cold data types. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a hot and cold data migration method.

[0147] Those skilled in the art will understand that Figure 12 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0148] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0149] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0150] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0151] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0152] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0153] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0154] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for transferring hot and cold data, characterized in that, The method includes: Retrieve customer information in response to trigger actions from the front-end application; The customer information is input into a pre-trained customer behavior prediction model to obtain the customer behavior prediction result; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data. The customer behavior prediction results are compared with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, as well as the predicted hot and cold data types of the associated data. Based on the actual hot and cold data types of the associated data and the predicted hot and cold data types, a migration strategy is generated for the associated data; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

2. The method according to claim 1, characterized in that, Before retrieving customer information in response to a trigger operation on the front-end application, the method further includes: Obtain the correlation between various data corresponding to the front-end application, the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data; A hot and cold data knowledge graph is constructed based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data.

3. The method according to claim 2, characterized in that, The step of constructing a hot and cold data knowledge graph based on customer behavior, various types of data from the front-end application, the relationships between various types of data corresponding to the front-end application, the relationships between customer behavior and various types of data from the front-end application, and the relationships between customer behavior and the predicted hot and cold data types of application data, includes: The customer is identified as the central node, and various data from the front-end application are identified as multiple radiating nodes; Based on the correlation between customer behavior and various data of the front-end application, and the correlation between customer behavior and the predicted hot and cold data types of application data, the edges between the central node and its adjacent radiating nodes are constructed. Based on the relationships between various data corresponding to the front-end application, edges are constructed between the adjacent radiating nodes and other radiating nodes in the direction away from the central node to obtain a hot and cold data knowledge graph.

4. The method according to claim 1, characterized in that, The step of generating a migration strategy for the associated data based on the actual hot and cold data types of the associated data and the predicted hot and cold data types includes: If the actual cold / hot data type of the associated data is the cold data and the predicted cold / hot data type is the hot data, then the migration strategy is determined to be to migrate the associated data to the first memory where the hot data is located. If the actual hot / cold data type of the associated data is hot data and the predicted hot / cold data type is cold data, then the migration strategy is determined to be to migrate the associated data to the second memory where the cold data is located.

5. The method according to claim 1, characterized in that, The method further includes: Collect real-time customer behavior and access data related to the real-time customer behavior; The real-time customer behavior is compared with the customer behavior prediction result to obtain the first comparison result; The access data is compared with the associated data to obtain a second comparison result; Based on the first comparison result and the second comparison result, the pre-trained customer behavior prediction model and the hot and cold data knowledge graph are optimized.

6. The method according to claim 5, characterized in that, The optimization process for the pre-trained customer behavior prediction model and the hot and cold data knowledge graph based on the first comparison result and the second comparison result includes: If the first comparison result is lower than the preset hit threshold, the pre-trained customer behavior prediction model is trained using the real-time customer behavior to obtain an optimized customer behavior prediction model. If the second comparison result is lower than the preset hit threshold, the hot and cold data knowledge graph is corrected based on the customer's real-time behavior and the access data to obtain the corrected hot and cold data knowledge graph.

7. A cold / hot data migration device, characterized in that, The device includes: The information acquisition module is used to acquire customer information in response to trigger operations on the front-end application; The behavior prediction module is used to input the customer information into a pre-trained customer behavior prediction model to obtain customer behavior prediction results; the pre-trained customer prediction model is obtained by training an initial neural network based on the customer's historical operation data. The type prediction module is used to compare the customer behavior prediction results with the hot and cold data knowledge graph to obtain the associated data corresponding to the customer behavior prediction results, and the predicted hot and cold data types of the associated data. The data migration module is used to generate a migration strategy for the associated data based on the actual hot and cold data types and the predicted hot and cold data types; the migration strategy is used to migrate the associated data to a first memory where the hot data is located, or to a second memory where the cold data is located.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images