Heuristic collaborative caching method based on graph convolutional neural network
Through the heuristic collaborative caching method based on graph convolutional neural network, the shortcomings of edge caching technology in content popularity prediction, cache placement strategy and storage resource utilization are solved, and efficient cache management and low-cost caching scheme are realized.
Patent Information
- Application Number
- CN202510347430.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-24
- Publication Date
- 2025-06-24
AI Technical Summary
The existing edge caching technology has shortcomings in content popularity prediction, cache placement strategy flexibility, storage resource utilization and latency energy consumption costs, especially in multi-user and multi-server scenarios.
The heuristic collaborative caching method based on graph convolution neural network is adopted to analyze the user-content relationship in user requests, build a knowledge graph, use graph convolutional neural network to predict the popularity of content, and formulate a heuristic cache content placement strategy to optimize the selection and placement of cached content.
It improves the cache hit rate, reduces the overall system cost, improves cache efficiency, and adapts to changes in multi-user and multi-server scenarios.
Smart Images
Figure CN120196561A_ABST
Abstract
Description
Technical Field
[0001] The present invention discloses a heuristic collaborative caching method based on a graph convolutional neural network, which relates to the field of edge computing caching technology. Background Art
[0002] With the rapid development of the mobile Internet, users' demand for high-quality and low-latency content is increasing day by day. To address this challenge, edge caching technology has emerged. By storing popular content at the edge of the network, such as base stations, the need to transmit data back to the core network is reduced, thereby reducing latency and alleviating the burden on the core network. In the scenario of multiple users and multiple servers, edge caching technology is particularly crucial. However, current edge caching technology still faces some deficiencies:
[0003] For example, the prediction of content popularity is inaccurate: Existing popularity prediction methods often rely on historical data or simple statistical models, which are inadequate when dealing with complex and changing user preferences. The cached content may not fully meet the actual needs of users, resulting in a low cache hit rate and serious resource waste;
[0004] The cache placement strategy lacks flexibility: Many existing cache placement strategies only consider local popular content and ignore the diversity of global content distribution, resulting in global content redundancy, that is, the same content is cached repeatedly by multiple base stations. This not only reduces the cache efficiency but also increases the system energy consumption.
[0005] The storage resource limitations of base stations are not fully considered: In actual deployment, the storage resources of base stations are limited. Current cache placement strategies often fail to make full use of the limited storage space, resulting in suboptimal selection and placement of cached content.
[0006] The latency and energy consumption costs are high: When the content requested by a user is not cached by the local base station, it needs to be obtained from other base stations or the core network, which increases the latency and energy consumption costs. Especially in the scenario of multiple users and multiple servers, these costs may be more significant.
[0007] The utilization of collaborative networks is lacking: Existing edge caching technologies often ignore the potential of collaborative networks, such as community base stations. Collaborative networks can optimize the cache placement strategy by sharing resources and information, further improving the cache efficiency and reducing latency. However, current technologies still have deficiencies in realizing this collaboration. Summary of the Invention
[0008] In view of the problems of the prior art, the present invention provides a heuristic collaborative caching method based on a graph convolutional neural network, which considers changes in user preferences and system cost factors in the scenario of multiple users and multiple servers, improves the cache hit rate, and reduces the overall system cost.
[0009] The specific solution proposed by the present invention is as follows:
[0010] The present invention provides a heuristic collaborative caching method based on a graph convolutional neural network, including:
[0011] Step 1: Based on the mobile edge computing system, analyze the user-content relationship according to the user request, and construct a knowledge graph according to the user-content relationship.
[0012] Step 2: According to the knowledge graph, use the content popularity prediction algorithm of the graph convolutional neural network to calculate the probability that the user is interested in the content, and obtain the predicted cached content set and the content popularity of each content.
[0013] Step 3: Formulate a heuristic caching content placement strategy, and place the cached content according to the heuristic caching content placement strategy:
[0014] Step 31: Initialize the cached content placement matrix and the capacity of each base station, create an auxiliary set, and initialize the minimum content number for the predicted cached content set.
[0015] Step 32: Determine whether there is a remaining capacity of a base station that meets the required capacity of the minimum content. If it meets, traverse the cached content set to initialize the optimal base station and the optimal cost.
[0016] Step 33: After initialization, traverse all base stations and determine whether the base station capacity meets the required size of the cached content. If it meets, calculate the collaborative caching cost and compare it with the optimal cost. When the collaborative caching cost is less than the optimal cost, update the optimal base station and the optimal cost. After the traversal ends, if the capacity of the optimal base station can cache the content, update the corresponding cached content to the cached content placement matrix; otherwise, add the cached content to the auxiliary set.
[0017] Step 34: Output the cached content in the cached content placement matrix to the corresponding base station.
[0018] Step 35: Output all the cached content in the auxiliary set to the community base station.
[0019] Furthermore, in step 1 of the heuristic collaborative caching method based on a graph convolutional neural network, constructing the knowledge graph includes:
[0020] Preprocess the data of the user-content relationship to obtain text data. Use the natural language processing method NLP to identify the entities in the text data, extract the relationship between the user and the content according to the entities, and design the structure of the knowledge graph according to the definitions of the entities and relationships to complete the construction of the knowledge graph.
[0021] Further, after step 33 of the heuristic cooperative caching method based on graph convolutional neural network is executed, the cache content set is re-initialized, and steps 32 - 33 are repeatedly executed until the base station cannot cache the content.
[0022] Further, in step 35 of the heuristic cooperative caching method based on graph convolutional neural network, the cache content in the auxiliary set is sorted in descending order according to content popularity, and then all are put into the community base station.
[0023] The present invention also provides a heuristic cooperative caching device based on graph convolutional neural network, including a graph spectrum management module, a content prediction module, and a placement strategy module.
[0024] The graph spectrum management module analyzes the user-content relationship based on the mobile edge computing system according to user requests, and constructs a knowledge graph according to the user-content relationship.
[0025] The content prediction module calculates the probability of a user's interest in content using the content popularity prediction algorithm of graph convolutional neural network based on the knowledge graph, and obtains the predicted cache content set and the content popularity of each content.
[0026] The placement strategy module formulates a heuristic cache content placement strategy, and places the cache content according to the heuristic cache content placement strategy:
[0027] Step 31: Initialize the cache content placement matrix and the capacity of each base station, create an auxiliary set, and initialize the minimum content number for the predicted cache content set.
[0028] Step 32: Determine whether there is a remaining capacity of a base station that meets the required capacity of the minimum content. If so, traverse the cache content set to initialize the optimal base station and the optimal cost.
[0029] Step 33: After initialization, traverse all base stations and determine whether the base station capacity meets the required size of the cache content. If so, calculate the cooperative caching cost and compare it with the optimal cost. When the cooperative caching cost is less than the optimal cost, update the optimal base station and the optimal cost. After the traversal ends, if the capacity of the optimal base station can cache the content, update the corresponding cache content to the cache content placement matrix; otherwise, add the cache content to the auxiliary set.
[0030] Step 34: Output the cache content in the cache content placement matrix to the corresponding base station.
[0031] Step 35: Output all the cache content in the auxiliary set to the community base station.
[0032] Further, the graph spectrum management module of the heuristic cooperative caching device based on graph convolutional neural network constructs a knowledge graph, including:
[0033] Preprocess the data of the user-content relationship to obtain text data. Use the natural language processing method NLP to identify entities in the text data, extract the relationship between the user and the content according to the entities, and design the structure of the knowledge graph according to the definitions of the entities and relationships to complete the construction of the knowledge graph.
[0034] Further, after the placement strategy module of the heuristic cooperative caching device based on the graph convolutional neural network executes step 33, re-initialize the cache content set, and repeat the execution of step 32-step 33 until the base station cannot cache the content.
[0035] Further, the placement strategy module of the heuristic cooperative caching device based on the graph convolutional neural network executes step 35, sorts the cache content in the auxiliary set in descending order according to the content popularity, and then puts all of them into the community base station.
[0036] The beneficial effects of the present invention are:
[0037] The present invention applies the technology of Graph Convolutional Neural Network (GCN). Through GCN, it captures the complex relationship between users and server content, constructs a knowledge graph, and adopts a heuristic placement strategy according to the knowledge graph to reduce the comprehensive cache cost. Through the heuristic strategy, it can quickly find a near-optimal cache placement scheme under limited network and storage resources, so as to minimize the comprehensive cost while ensuring the user experience. The placement strategy of the present invention is not only applicable to the current multi-user multi-server scenario, but also has good adaptability and scalability. With the continuous change of the network environment and the increasing diversification of user needs, it can adapt to new scenarios and needs by adjusting the model parameters of GCN and the relevant parameters of the heuristic optimization algorithm. In addition, this strategy is also easy to combine with other cache management strategies to form a more perfect and efficient cache management system, providing strong technical support for the future network development. Brief Description of the Drawings
[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to these drawings.
[0039] Figure 1 It is a schematic flowchart of the method of the present invention.
[0040] Figure 2 It is a schematic diagram of the user-content relationship.
[0041] Figure 3 It is a schematic diagram of the user-content mapping. Specific implementation manners
[0042] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, so that those skilled in the art can better understand the present invention and be able to implement it, but the exemplified embodiments do not limit the present invention.
[0043] Embodiment 1
[0044] The present invention provides a heuristic collaborative caching method based on a graph convolutional neural network, including:
[0045] Step 1: Analyze the user-content relationship based on a mobile edge computing system according to user requests, and construct a knowledge graph according to the user-content relationship. When constructing the knowledge graph, it may include:
[0046] Preprocess the data of the user-content relationship to obtain text data, use the natural language processing method NLP to identify entities in the text data, extract the relationship between the user and the content according to the entities, and design the structure of the knowledge graph according to the definitions of the entities and relationships to complete the construction of the knowledge graph.
[0047] Step 2: According to the knowledge graph, use the content popularity prediction algorithm of the graph convolutional neural network to calculate the probability that the user is interested in the content, and obtain the predicted cached content set and the content popularity of each content.
[0048] Among them, when using the content popularity prediction algorithm of the graph convolutional neural network to calculate the probability that the user is interested in the content, the following can be referred to for the calculation execution code process: ug is the user, l is the cached content, V is the knowledge graph mapping entity, and r is the relationship between entities.
[0049]
[0050] The first line represents looping through the input user-content pairs; the second to third lines represent iteratively calculating the neighborhood representation of entity e in each layer; the fourth to tenth lines represent obtaining the aggregated vector after aggregating H; the eleventh to fifteenth lines represent calculating the probability that the user is interested in the cached content to obtain the cached content set and the content popularity of each content; the sixteenth to twenty-fourth lines are the specific calculation process for iteratively obtaining the receptive field of content l.
[0051] Step 3: Develop a heuristic caching content placement strategy and place the cached content according to the heuristic caching content placement strategy:
[0052] Step 31: Initialize the cached content placement matrix and the capacity of each base station and create an auxiliary set, and initialize the minimum content number for the predicted cached content set.
[0053] Step 32: Determine whether there is remaining capacity of the base station that meets the required capacity of the minimum content. If it meets, traverse the cache content set to initialize the optimal base station and the optimal cost.
[0054] After the initialization is completed in Step 33, traverse all base stations and determine whether the base station capacity meets the required size of the cached content. If it meets, calculate the collaborative caching cost and compare it with the optimal cost. When the collaborative caching cost is less than the optimal cost, update the optimal base station and the optimal cost. After the traversal ends, if the capacity of the optimal base station can cache the content, update the corresponding cached content to the cache content placement matrix; otherwise, add the cached content to the auxiliary set, re-initialize the cache content set, and repeat Steps 32 - 33 until the base station cannot cache the content.
[0055] Step 34: Output the cached content in the cache content placement matrix to the corresponding base station.
[0056] Step 35: Sort the cached content in the auxiliary set in descending order according to the content popularity, and output all the cached content to the community base station. The calculation and execution process of the code is as follows:
[0057]
[0058]
[0059] Lines 1 - 7 are for initialization work. First, initialize the cache content placement matrix and the base station capacity and create the auxiliary set Ξ. Then, obtain the cache content set obtained by the content popularity prediction algorithm based on the graph convolutional neural network and the user's preference probability for the content, and initialize the minimum content number. Lines 8 - 23 are for the calculation process. First, determine whether there is remaining capacity of the base station that meets the required capacity of the minimum content. If it meets, traverse the content set to initialize the optimal base station and the optimal cost. After all the initialization work is completed, traverse the base station set and determine whether the base station capacity meets the required size of the cached content. If it meets, calculate the collaborative caching cost and compare it with the optimal cost. When the cost is less than the optimal cost, update the optimal base station and the optimal cost. After the traversal ends, if the optimal base station capacity can cache the content, update the placement matrix; otherwise, add the content to the auxiliary matrix. After the content set traversal ends, sort the cached content in the auxiliary set in descending order according to its content popularity, and re-initialize the cache content set, repeating lines 9 - 22 until the base station cannot meet the caching condition. Finally, output the content placement matrix and the auxiliary set.
[0060] Embodiment 2
[0061] The present invention also provides a heuristic collaborative caching device based on a graph convolutional neural network, including a graph spectrum management module, a content prediction module, and a placement strategy module.
[0062] The atlas management module analyzes the user-content relationship based on the mobile edge computing system according to the user's request, and constructs a knowledge graph based on the user-content relationship.
[0063] The content prediction module calculates the probability that the user is interested in the content using the content popularity prediction algorithm of the graph convolutional neural network based on the knowledge graph, and obtains the predicted cached content set and the content popularity of each content.
[0064] The placement strategy module formulates a heuristic caching content placement strategy and places the cached content according to the heuristic caching content placement strategy:
[0065] Step 31: Initialize the cached content placement matrix and the capacity of each base station, create an auxiliary set, and initialize the minimum content number for the predicted cached content set.
[0066] Step 32: Determine whether there is a remaining capacity of a base station that meets the required capacity of the minimum content. If it is satisfied, traverse the cached content set to initialize the optimal base station and the optimal cost.
[0067] After the initialization is completed, traverse all base stations and determine whether the base station capacity meets the required size of the cached content. If it is satisfied, calculate the collaborative caching cost and compare it with the optimal cost. When the collaborative caching cost is less than the optimal cost, update the optimal base station and the optimal cost. After the traversal ends, if the capacity of the optimal base station can cache the content, update the corresponding cached content to the cached content placement matrix; otherwise, add the cached content to the auxiliary set.
[0068] Step 34: Output the cached content in the cached content placement matrix to the corresponding base station.
[0069] Step 35: Output all the cached content in the auxiliary set to the community base station.
[0070] Regarding the information interaction and execution process among the above-mentioned modules in the device, since they are based on the same concept as the method embodiment of the present invention, the specific content can be referred to the description in the method embodiment of the present invention, and will not be elaborated here.
[0071] Similarly, the device of the present invention utilizes the technology of Graph Convolutional Neural Network (GCN). By means of GCN, the complex relationships between users and server content are captured to construct a knowledge graph, and a heuristic placement strategy is adopted according to the knowledge graph to reduce the overall cache cost. Through the heuristic strategy, a nearly optimal cache placement scheme can be quickly found under limited network and storage resources, so as to minimize the overall cost while ensuring the user experience. The placement strategy of the present invention is not only applicable to the current multi-user and multi-server scenario, but also has good adaptability and scalability. With the continuous change of the network environment and the increasing diversification of user requirements, the model parameters of GCN and the relevant parameters of the heuristic optimization algorithm can be adjusted to adapt to new scenarios and requirements. In addition, this strategy is also easy to be combined with other cache management strategies to form a more perfect and efficient cache management system, providing strong technical support for the future development of the network.
[0072] It should be noted that not all steps and modules in the above-mentioned processes and device structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted according to needs. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities respectively, or some components in multiple independent devices may be jointly implemented.
[0073] The above-mentioned embodiments are only preferred embodiments given to fully illustrate the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or transformations made by those skilled in the art on the basis of the present invention are all within the protection scope of the present invention. The protection scope of the present invention is subject to the claims.
Claims
1. A heuristic collaborative caching method based on graph convolutional neural network, characterized by include: Step 1: Analyze the user-content relationship based on the user request based on the mobile edge computing system, and build a knowledge graph based on the user-content relationship. Step 2: Based on the knowledge graph, the content popularity prediction algorithm of the graph convolutional neural network is used to calculate the probability of users being interested in the content, and the predicted cache content set and the content popularity of each content are obtained. Step 3: Develop a heuristic cache content placement strategy and place cache content according to the heuristic cache content placement strategy: Step 31: Initialize the cache content placement matrix and the capacity of each base station and create an auxiliary set, initialize the minimum content number for the predicted cache content set, Step 32: Determine whether there is a base station with a remaining capacity that meets the minimum content capacity. If so, traverse the cache content set to initialize the optimal base station and optimal cost. Step 33: After the initialization is completed, all base stations are traversed to determine whether the base station capacity meets the required size of the cached content. If so, the collaborative cache cost is calculated and compared with the optimal cost. When the collaborative cache cost is less than the optimal cost, the optimal base station and the optimal cost are updated. At the end of the traversal, if the capacity of the optimal base station can cache the content, the corresponding cached content is updated to the cache content placement matrix, otherwise the cached content is added to the auxiliary set. Step 34: Output the cache content of the cache content placement matrix to the corresponding base station, Step 35: Output all cache contents of the auxiliary set to the community base station.
2. According to claim 1, a heuristic collaborative caching method based on graph convolutional neural network is characterized in that In step 1, we build a knowledge graph, including: Preprocess the data on user-content relationships to obtain text data, use the natural language processing method NLP to identify entities in the text data, and extract the relationship between users and content based on the entities. Based on the definitions of entities and relationships, design the structure of the knowledge graph to complete the knowledge graph construction.
3. According to claim 1, a heuristic collaborative caching method based on graph convolutional neural network is characterized in that After step 33 is executed, the cache content set is reinitialized, and steps 32 to 33 are repeatedly executed until the base station is unable to cache content.
4. The heuristic collaborative caching method based on graph convolutional neural network according to claim 1 is characterized in that In step 35, the cached contents in the auxiliary set are sorted in descending order according to the content popularity, and then all are put into the community base station.
5. A heuristic collaborative caching device based on graph convolutional neural network, characterized in that It includes graph management module, content prediction module and placement strategy module. The graph management module analyzes the user-content relationship based on the mobile edge computing system according to the user request, and builds a knowledge graph based on the user-content relationship. The content prediction module calculates the probability of users being interested in content based on the knowledge graph and uses the content popularity prediction algorithm of the graph convolutional neural network, and obtains the predicted cache content set and the content popularity of each content. The placement strategy module formulates a heuristic cache content placement strategy and places cache content according to the heuristic cache content placement strategy: Step 31: Initialize the cache content placement matrix and the capacity of each base station and create an auxiliary set, initialize the minimum content number for the predicted cache content set, Step 32: Determine whether there is a base station with a remaining capacity that meets the minimum content capacity. If so, traverse the cache content set to initialize the optimal base station and optimal cost. Step 33: After the initialization is completed, all base stations are traversed to determine whether the base station capacity meets the required size of the cached content. If so, the collaborative cache cost is calculated and compared with the optimal cost. When the collaborative cache cost is less than the optimal cost, the optimal base station and the optimal cost are updated. At the end of the traversal, if the capacity of the optimal base station can cache the content, the corresponding cached content is updated to the cache content placement matrix, otherwise the cached content is added to the auxiliary set. Step 34: Output the cache content of the cache content placement matrix to the corresponding base station, Step 35: Output all cache contents of the auxiliary set to the community base station.
6. According to claim 5, a heuristic collaborative caching device based on graph convolutional neural network is characterized in that the graph The management module builds the knowledge graph, including: Preprocess the data on user-content relationships to obtain text data, use the natural language processing method NLP to identify entities in the text data, and extract the relationship between users and content based on the entities. Based on the definitions of entities and relationships, design the structure of the knowledge graph to complete the knowledge graph construction.
7. The heuristic collaborative caching device based on graph convolutional neural network according to claim 5 is characterized in that After the placement strategy module executes step 33, the cache content set is reinitialized, and steps 32 to 33 are repeatedly executed until the base station cannot cache the content.
8. The heuristic collaborative caching device based on graph convolutional neural network according to claim 5 is characterized by: The placement strategy module executes step 35 to sort the cached contents in the auxiliary set in descending order according to the content popularity, and then puts all of them into the community base station.