Account processing method based on computing network energy demand and computer program product

Abstract

The invention discloses an account processing method based on computing network energy requirements and a computer program product. The method relates to the field of data processing, and comprises the following steps: determining resource demand types respectively corresponding to a plurality of accounts with network energy demands, the network energy demands being used for indicating that the corresponding accounts have computing power resources, network resources and energy resources; determining a plurality of initial clustering centers for clustering the plurality of accounts based on the resource demand types corresponding to the plurality of accounts; clustering the plurality of accounts based on the plurality of initial clustering centers to obtain a clustering result; and based on the clustering result, determining a classification result of the plurality of accounts. According to the method and the device, the technical problem of low classification result accuracy caused by adopting a clustering mode to classify the energy demand accounts based on a randomly selected initial clustering center in related technologies is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an account processing method and a computer program product based on computing network energy demand. Background Art

[0002] In recent years, computing power has gradually become a new productivity. With the widespread use of digital technologies such as artificial intelligence, blockchain, and the Internet of Things, and the large-scale deployment of computing power infrastructure such as artificial intelligence data centers and supercomputing data centers, users' computing network needs are growing, and green and low-carbon awareness is deepening. It is difficult to significantly improve the comprehensive economic benefits of resource utilization by optimizing computing network energy resources on the supply side. Considering that users are an important source of economic profits for enterprises, they have an important impact on the overall development of enterprises. Therefore, it is necessary to start from the user side, clarify user needs, and design a clustering method based on the computing network energy user side.

[0003] Given the current situation of diverse user needs, large data scale, and slow clustering convergence, it is challenging to build a clustering method with high user satisfaction and high accuracy. Clustering methods in related technologies are mostly based on randomly generated initial cluster centers and use K-means clustering (k-means) method to classify users, which makes it difficult to select the initial distance center, and thus leads to low accuracy of the results obtained by classifying the computing network energy demand accounts.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide an account processing method based on network energy demand and a computer program product, so as to at least solve the technical problem in the related art that the energy demand accounts are classified by clustering method based on randomly selected initial cluster centers, resulting in low accuracy of classification results.

[0006] According to one aspect of an embodiment of the present invention, there is provided an account processing method based on computing network energy demand, comprising: determining resource demand types respectively corresponding to a plurality of accounts having computing network energy demand, wherein the computing network energy demand is used to indicate that the corresponding account has computing power resources, network resources, and energy resources; based on the resource demand types respectively corresponding to the plurality of accounts, determining a plurality of initial clustering centers for clustering the plurality of accounts; based on the plurality of initial clustering centers, clustering the plurality of accounts to obtain clustering results; and based on the clustering results, determining classification results for the plurality of accounts.

[0007] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing any one of the account processing methods based on computing network energy demand.

[0008] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of any one of the methods for account processing based on computing network energy demand.

[0009] In an embodiment of the present invention, by determining the resource demand types respectively corresponding to a plurality of accounts having computing network energy demand, wherein the computing network energy demand is used to indicate that the corresponding account has computing power resources, network resources, and energy resources; based on the resource demand types respectively corresponding to the plurality of accounts, determining a plurality of initial clustering centers for clustering the plurality of accounts; based on the plurality of initial clustering centers, clustering the plurality of accounts to obtain clustering results; based on the clustering results, determining the classification results of the plurality of accounts, thereby achieving the purpose of determining a plurality of initial clustering centers based on the resource demand type of each energy demand account, clustering the accounts based on the determined initial clustering centers, and obtaining clustering results that better meet the user's resource demand, thereby achieving the technical effect of improving the accuracy of account clustering results, and further solving the technical problem of low accuracy of classification results in the related art based on randomly selected initial clustering centers and adopting clustering method to classify energy demand accounts. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0011] Figure 1 is a flow chart of an account processing method based on computing network energy demand according to an embodiment of the present invention;

[0012] Figure 2 is a schematic diagram of an optional computing network energy user demand mapping architecture according to an embodiment of the present invention;

[0013] Figure 3 is a flow chart of an optional account processing method based on computing network energy demand according to an embodiment of the present invention;

[0014] Figure 4 This is an optional flow chart of tagging network energy user indicators according to an embodiment of the present invention;

[0015] Figure 5 is an optional initial cluster center screening flow chart according to an embodiment of the present invention;

[0016] Figure 6 is an optional flow chart of clustering users of computing network according to an embodiment of the present invention;

[0017] Figure 7 It is a schematic diagram of an account processing device based on computing network energy demand according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] First, to facilitate understanding of the embodiments of the present invention, some terms or nouns involved in the present invention are explained below:

[0021] K-means clustering (k-means) algorithm is an unsupervised learning method used to divide samples in a data set into different categories. In k-means clustering, you first need to specify the number of categories k to be divided, and then the algorithm iteratively assigns data points to the nearest category center and updates the center position of each category until the convergence condition is reached. The final result is to divide the data set into k categories so that each data point belongs to the nearest category center.

[0022] The K-medoids algorithm is a clustering algorithm based on centroids, similar to the k-means algorithm, but the difference is that the k-medoids algorithm selects actual data points as centroids instead of calculated averages. In the k-medoids algorithm, the centroids are called medoids, and they represent representative samples of each cluster. The k-medoids algorithm performs clustering by minimizing the sum of the distances from data points to the centroids (medoids).

[0023] Sparrow Search Algorithm is an optimization algorithm based on bird foraging behavior, simulating the search process of sparrows when looking for food. Sparrow Search Algorithm is a heuristic algorithm, mainly used to solve optimization problems, such as function optimization, combinatorial optimization, machine learning and other fields.

[0024] The adaptive t distribution method is a method used in clustering algorithms to estimate the distance between data points. It is based on the t distribution and adapts to the distribution of data by adjusting the degrees of freedom of the distribution.

[0025] In recent years, computing power has gradually become a new productivity. With the widespread use of digital technologies such as artificial intelligence, blockchain, and the Internet of Things, and the large-scale deployment of computing power infrastructure such as artificial intelligence data centers and supercomputing data centers, users' computing network needs are growing, and green and low-carbon awareness is deepening. It is difficult to significantly improve the comprehensive economic benefits of resource utilization by optimizing computing network energy resources on the supply side. Considering that users are an important source of economic profits for enterprises, they have an important impact on the overall development of enterprises. Therefore, it is necessary to start from the user side, clarify user needs, and design a clustering method based on the computing network energy user side.

[0026] Given the current situation of diverse user needs, large data scale, and slow clustering convergence, it is challenging to construct a clustering method with high user satisfaction and high accuracy. Clustering methods in related technologies are mostly based on randomly generated initial cluster centers and use the k-means method to classify users, which makes it difficult to select the initial distance center. The accurate implementation of the computing network energy user clustering method is of great significance for optimizing resource allocation, improving service quality, promoting sustainable energy development, reducing environmental impact, and promoting the intelligent development of computing networks. However, the computing network energy user clustering methods in related technologies mostly start from the supply side and ignore the mapping relationship between task applications and user needs, resulting in low user satisfaction.

[0027] In response to the above problems, an embodiment of the present invention provides a method embodiment of account processing based on network energy demand. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] Figure 1 is a flow chart of an account processing method based on network energy demand according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0029] Step S102, determining resource demand types corresponding to multiple accounts with computing network energy demand, wherein the computing network energy demand is used to indicate that the corresponding account has computing power resources, network resources, and energy resources;

[0030] In step S102, each of the multiple accounts corresponds to a computing network user, which is used to indicate a user with computing network resources (i.e., computing power resources, network resources, and energy resources) requirements. By determining the resource requirement type of the account, accurate mapping of account requirements can be achieved to more accurately understand the specific preferences or requirements of each account when using computing network resources.

[0031] Optionally, this resource demand type is used to indicate the different requirements of the corresponding account for computing resources, network resources, and energy resources, which may include but are not limited to high-sensitivity green computing demand, low-sensitivity green computing demand, high-sensitivity green network demand, and low-sensitivity green network demand. Among them, high-sensitivity green computing demand requires a large amount of computing resources to support, and has strict requirements on energy consumption and carbon emissions, which may include but are not limited to model training, scientific computing, genomics processing, meteorological simulation, etc.; low-sensitivity green computing demand requires a large amount of computing resources to support, but has relatively low requirements on energy consumption and carbon emissions, including model reasoning, image recognition, cloud computing services, game graphics rendering, movie special effects production, etc.; high-sensitivity green network demand has high requirements on network bandwidth, stability and latency, and also has a high concern for energy consumption and carbon emissions, which may include but are not limited to social 3.0, e-commerce, online banking and financial services, etc.; low-sensitivity green network demand has high requirements on network bandwidth, stability and latency, but has relatively low requirements on energy consumption and carbon emissions, which may include but are not limited to reservation systems, virtual reality applications, etc.

[0032] In an optional embodiment, determining resource demand types corresponding to multiple accounts with computing network energy requirements includes: obtaining resource indicators corresponding to the multiple accounts, wherein the resource indicators include computing power indicators corresponding to computing power resources, network indicators corresponding to network resources, and energy indicators corresponding to energy resources; based on the resource indicators corresponding to the multiple accounts, determining the resource demand types corresponding to the multiple accounts.

[0033] Optionally, the computing power indicators, network indicators and energy indicators of multiple accounts provide specific requirements and preference information for the accounts in using computing power, network and energy resources. For example, the computing power indicators may include but are not limited to the utilization rate, main frequency, turbo frequency, etc. of the central processing unit (CPU), where the main frequency refers to the basic operating frequency of the CPU under normal working conditions without any acceleration or overclocking, and the turbo frequency refers to the frequency that the CPU can dynamically increase under certain conditions; the network indicators may include but are not limited to network bandwidth (such as upstream and downstream bandwidth), latency, packet loss rate, etc.; the energy indicators may include but are not limited to energy efficiency level, energy efficiency ratio (i.e., the proportion of green energy use), carbon emissions, etc. Based on the collected resource indicators, the resource requirements of the account are mapped to the predefined resource requirement types. This mapping helps to understand the specific needs of each account more accurately, so that resources can be allocated more reasonably in the subsequent clustering process, improve the accuracy of the clustering algorithm, and make the subsequent clustering results closer to reality.

[0034] Optionally, multiple accounts can correspond to the same task type or different task types. In the use of computing network energy resources, different types of tasks have different requirements for computing power, network, and energy resources. For example, image processing may require more GPU computing, while large-scale data analysis may rely more on CPU and memory. The user's sensitivity to green energy also determines how resources are used. Therefore, a mapping framework for computing network energy user needs can be constructed to achieve accurate matching of resources required for tasks of different types and scenarios, and to fully understand the actual needs of users. Specifically: starting from the user side, we can fully understand the relationship between "tasks-user needs-resources" and build a three-dimensional computing network energy user demand mapping system including task layer, user demand layer, and indicator layer. Clearly define the user demand layer, including high-sensitivity green computing strength demand, low-sensitivity green computing strength demand, high-sensitivity green network strength demand, and low-sensitivity green network strength demand. The completed computing network energy user demand mapping system, Figure 2 It is a schematic diagram of an optional computing network energy user demand mapping architecture according to an embodiment of the present invention.

[0035] like Figure 2As shown, the computing network energy task types can be divided into four categories: AI type, big data type, distributed computing type, and graphics rendering type. Among them, AI type service scenarios include image recognition, model training, model reasoning, and social 3.0; big data type service scenarios include e-commerce, online banking and financial services, and reservation systems; distributed computing type service scenarios include scientific computing, meteorological simulation, cloud computing services, and genome processing; graphics rendering type service scenarios include game graphics rendering, movie special effects production, and virtual reality applications. On this basis, according to user needs, the resource demand types can be subdivided into high-sensitivity green computing demand, low-sensitivity green computing demand, high-sensitivity green network demand, and low-sensitivity green network demand.

[0036] On this basis, the user's computing power indicators, network indicators, and energy indicators are clarified. Among them, the computing power indicators are used to evaluate the computing power resources required by the user, which can include but are not limited to the main frequency, turbo frequency, CPU usage, etc.; the network indicators are used to evaluate the network resources required by the user, which can include but are not limited to bandwidth, latency, packet loss rate, etc.; the energy indicators are used to evaluate the energy resources required by the user, which can include but are not limited to performance level, energy efficiency ratio, carbon emissions, etc.

[0037] In an optional embodiment, based on the resource indicators corresponding to the multiple accounts respectively, the resource demand types corresponding to the multiple accounts are determined, including: obtaining the resource demand type of any account among the multiple accounts in the following manner: determining the indicator score of any account based on the resource indicators of any account, wherein the indicator score includes a computing power indicator score, a network indicator score, and an energy indicator score; determining the degree of match between any account and the multiple resource demand types based on preset indicator scores corresponding to the multiple resource demand types and the indicator score of any account, wherein the preset indicator score includes a preset computing power indicator score, a preset network indicator score, and a preset energy indicator score; determining the resource demand type of any account based on the degree of match between any account and the multiple resource demand types; and obtaining the resource demand types corresponding to the multiple accounts by adopting the method of obtaining the resource demand type of any account.

[0038] It can be understood that by quantifying the specific needs of each account in terms of computing power, network and energy through indicator scoring, and calculating the matching degree between the preset indicator score and the actual account indicator score, users can be more accurately classified into different resource demand types, such as high-sensitivity green computing strength demand, low-sensitivity green computing strength demand, high-sensitivity green network strength demand, low-sensitivity green network strength demand, etc. The greater the matching degree, the more consistent the account's resource demand characteristics are with the corresponding resource demand type, thereby more accurately determining the resource demand type of each account.

[0039] In an optional embodiment, when there are multiple computing power indicators, multiple network indicators, and multiple energy indicators, the indicator score of any account is determined based on the resource indicators of any account, including: determining the weight values ​​corresponding to the multiple computing power indicators respectively; performing weighted calculation on the multiple computing power indicators and the weight values ​​corresponding to the multiple computing power indicators respectively to obtain a computing power indicator score; determining the weight values ​​corresponding to the multiple network indicators respectively; performing weighted calculation on the multiple network indicators and the weight values ​​corresponding to the multiple network indicators respectively to obtain a network indicator score; determining the weight values ​​corresponding to the multiple energy indicators respectively; performing weighted calculation on the multiple energy indicators and the weight values ​​corresponding to the multiple energy indicators respectively to obtain an energy indicator score.

[0040] Optionally, by weighted calculation of multiple computing power indicators, network indicators and energy indicators, the user's demand for different resources can be comprehensively evaluated. This comprehensive evaluation can more comprehensively reflect the user's computing network energy resource needs and avoid the one-sidedness and inaccuracy that may be caused by a single indicator. The weight value is determined based on the importance of the user's specific needs. For example, for a user who needs to perform large-scale image processing, the weight of the GPU computing power indicator may be higher, while for an online financial service user who needs a stable network connection, the weight of the network stability and latency indicators may be assigned a higher value. This personalized demand mapping helps to more accurately identify and classify users, thereby providing services that better meet user needs. The computing power, network and energy indicator scores obtained by weighted calculation can be used as a basis for resource optimization configuration. For example, users with high scores may have priority to obtain better resources, while users with low scores may need to accept reasonable resource allocation or optimized use strategies, which helps to improve resource utilization efficiency and overall service level. In the clustering algorithm, the selection of the initial cluster center has an important impact on the clustering results.

[0041] It should be noted that the clustering methods in the related art have deficiencies in both initial center selection and local optimal control, which further reduces the rationality and efficiency of clustering. For example, the optimization ability of the k-means-based user clustering method is limited, and it is easy to fall into the local optimal solution, resulting in the lack of efficiency of clustering convergence. This embodiment introduces multi-label information and resource indicator scoring, takes into account more comprehensive information when selecting the initial clustering center, reduces the risk of the algorithm falling into the local optimal, and improves the accuracy and reliability of the clustering results.

[0042] Optionally, the corresponding resource demand type can be used as a label for each account. The matching degree between any account and multiple resource demand types can be characterized by the similarity between user resources and resource demand types (such as cosine similarity). When determining the resource demand type label for each account, considering the coarse-grained classification of the computing network energy user demand mapping and the diversity of computing network energy indicators, if large-scale account data is not labeled, the clustering effect and accuracy will be reduced. Therefore, for the constructed computing network energy user demand mapping system, all account data are calculated for the computing network energy indicator matching degree with four clearly defined resource demand types.

[0043] Specifically, in the calculation of the matching degree between user resources and resource demand types (hereinafter referred to as user matching degree), it is known that the user scale U = {u i}, i = 1, 2, ..., n, the user's computing power index score, network index score, and energy index score are u i = {u i-com ,u i-net ,u i-ene}, the ideal values ​​of the four types of user needs defined (i.e., the preset indicator scores) are high sensitivity green strong demand R1 = {R 1-com ,R 1-net ,R 1-ene}、Low sensitivity green computing strong demand R2={R 2-com ,R 2-net ,R 2-ene}、Highly sensitive green network strength requirement R3={R 3-com ,R 3-net ,R 3-ene}、Low sensitivity green network strength requirement R4={R 4-com ,R 4-net ,R 4-ene}, where u i-com Indicates the computing power index score of any account, u i-net represents the network indicator score of any account, u i-ene Represents the energy index score of any account; R j (j=1, 2, 3, 4) represents the preset indicator score of each resource demand type, i is the identifier of any account; R j-com (j=1,2,3,4) represents the preset computing power index score of each resource demand type, and j=1,2,3,4 represents the identifier of each resource demand type; R j-net (j=1, 2, 3, 4) represents the preset network indicator score of each resource demand type; R j-ene (j=1, 2, 3, 4) represents the preset energy index score of each resource demand type.

[0044] Among them, the user's computing network energy index (i.e., resource index) can be further standardized and multi-type indexes can be weighted to obtain the corresponding index score. Taking the user's computing power resource demand as an example, the computing power index score can be obtained in the following way:

[0045]

[0046] Among them, u i-com-l (l=1,2,…,m) represents any one of the multiple computing power indicators, ω l (l=1,2,…,m) represents the weight of any computing power indicator, m represents the total number of multiple computing power indicators, and

[0047] The matching degree between any user resource and any resource requirement type sim i,j , that is, the matching degree between any account and any resource demand type, is obtained as follows:

[0048]

[0049] in, Indicates the weight value of computing resource demand. represents the network resource demand weight, represents the energy resource demand weight, and

[0050] Based on the calculation of the matching degree between multiple user resources and any resource demand type obtained by the above method, the matching degrees between multiple user resources and any resource demand type are sorted, and the best matching degree (i.e., the best matching degree) is selected and recorded as sim best , complete the calculation and sorting of user matching degrees.

[0051] Optionally, the network energy indicator matching degree of all account data (i.e., multiple accounts) and four clearly defined resource demand types can be calculated and sorted; secondly, the user demand type can be determined by comparing the change in the matching degree of each account relative to the four resource demand types with the threshold setting, wherein the resource demand type corresponding to each account can be one or more, and the one or more resource demand types can be used as labels for the corresponding accounts, thereby completing the determination of the resource demand type labels for each account.

[0052] In an optional embodiment, the resource demand type of any account is determined based on the matching degree between any account and multiple resource demand types, including: obtaining the matching degree change of any account relative to any resource demand type among multiple resource demand types in the following manner: determining a maximum value among the matching degrees between multiple accounts and any resource demand type as the optimal matching degree; obtaining the matching degree change of any account relative to any resource demand type based on the matching degree between any account and any resource demand type and the optimal matching degree, wherein the matching degree change is used to indicate the degree of change of the matching degree between the corresponding account and the resource demand type relative to the corresponding optimal matching degree; obtaining the matching degree change of any account's resource indicators relative to multiple resource demand types by adopting the method of obtaining the matching degree change of any account relative to any resource demand type; determining the resource demand type of any account based on the matching degree change of any account's resource indicators relative to multiple resource demand types.

[0053] It can be understood that by calculating the change in the matching degree between each account and different resource demand types, the degree of match between the user account demand and the resource type can be quantified, which not only considers the absolute demand of the account, but also pays attention to its relative position among all demand types, which helps to more accurately identify the account's preference for specific resources. By introducing the optimal matching degree as a benchmark, it helps to determine the priority of account demand types. By comparing the matching degree between account demand and resource type with the optimal matching degree, it is possible to more clearly identify which resource types are optimal for a specific account, which helps to accurately screen out the type that best meets the account demand from multiple similar resource demand types, avoiding waste of resources and mismatched allocation.

[0054] Optionally, based on the change in the matching degree of the resource indicators of any account relative to a plurality of resource demand types, the resource demand type of any account is determined by threshold comparison, that is, the change in the matching degree of the resource indicators of any account relative to a plurality of resource demand types is compared with a preset threshold. If it is less than or equal to the threshold, it is determined that the any account meets the any resource demand type, otherwise it is determined that the any account does not meet the any demand type; the above operations are performed for a plurality of resource demand types respectively, and finally the resource demand type to which the any account belongs is determined. It can be seen that, based on the above method, each account can meet one or more resource demand types. For accounts that meet multiple resource demand types, multiple resource demand types are used as labels for the account; for accounts that do not meet multiple resource demand types, the resource demand type whose corresponding matching degree change is closest to the preset threshold is determined as the resource demand type matched by the account.

[0055] Optionally, by comparing the matching values ​​of the four types of requirements with the threshold settings, one or more labels are added to the user data to complete the user requirement type labeling. In user labeling, based on the calculation and sorting of user matching, the matching change of any account relative to any resource requirement type (i.e., matching change) is determined. The calculation formula is as follows:

[0056]

[0057] Among them, cha i Indicates the change in matching degree of any account relative to any resource requirement type, sim i,j Indicates the matching degree between any user resource and any resource demand type, sim best Indicates the best match.

[0058] By setting the deviation threshold α, the resource demand type of each user is determined, and the resource demand type is used as the label of the corresponding user to achieve labeling of user data. The judgment rule for each user is determined as follows: i ≤α, add user data tag type, for those that meet cha i All user resource demand types within the range of ≤α are defined as user labels (i.e. resource demand type labels). Otherwise, the user demand type with the best matching degree is defined as the only user label. Finally, the user index labeling of the computing network is completed, providing a basis for the selection of the initial clustering center.

[0059] Step S104, determining a plurality of initial clustering centers for clustering the plurality of accounts based on the resource demand types respectively corresponding to the plurality of accounts;

[0060] In step S104, the initial cluster center is determined based on the resource demand type of the account, which can ensure that the starting point of the cluster analysis is based on actual user needs, thereby improving the rationality and efficiency of clustering.

[0061] In an optional embodiment, based on the resource demand types corresponding to the multiple accounts, multiple initial clustering centers for clustering the multiple accounts are determined, including: using the resource demand types corresponding to the multiple accounts as labels, clustering the multiple accounts using a K-center point clustering algorithm to obtain multiple initial clustering centers.

[0062] It should be noted that the computing network user clustering method in the related art ignores the importance of label information for the selection of the initial center of the cluster, and has limitations when clustering large-scale data, resulting in a lack of rationality in the selection of the initial center. This embodiment uses the resource demand type as a clustering label, which directly reflects the resource demand characteristics of the account, thereby ensuring that the selection of the initial clustering center is more reasonable. The K-medoids algorithm selects actual data points as centers, which means that even if there are outliers or noise in the data set, these center points are more likely to represent the characteristics of most data, thereby improving the stability of clustering. In scenarios where the computing power, network and energy demand distribution may be uneven or form complex shapes, the k-medoids algorithm can better handle these non-spherical clusters, thereby improving the accuracy and practicality of clustering. By clustering accounts according to their resource demand types, computing power, network bandwidth and energy resources can be allocated more effectively to meet the specific needs of accounts of different categories and achieve optimal allocation of resources.

[0063] It should be noted that the k-medoids algorithm is an improved clustering algorithm. Compared with the k-means algorithm, it selects actual data points as the centers of clusters, which makes the calculation of cluster centers more intuitive and reasonable, especially when dealing with non-spherical clusters or data sets containing noise and outliers.

[0064] As an optional embodiment, resource demand types corresponding to multiple accounts are used as labels, and a K-center point clustering algorithm is used to cluster the multiple accounts to obtain multiple initial cluster centers, including: based on the resource demand types corresponding to the multiple accounts and the resource indicators corresponding to the multiple accounts, the multiple accounts are divided into multiple clusters, wherein the number of the multiple clusters is less than the number of the multiple accounts; the center points corresponding to the multiple clusters are determined; for any cluster among the multiple clusters, the distances between the points included in any cluster and the corresponding center points are determined; based on the distances between the points included in any cluster and the corresponding center points, the multiple clusters are updated to obtain multiple updated clusters; the above operations are repeated until a preset number of iterations is reached; the center points of the updated multiple clusters obtained when the preset number of iterations is reached are used as multiple initial cluster centers.

[0065] Optionally, in the selection of the initial cluster center, firstly, considering that multi-label user data (i.e., one computing network user may correspond to multiple resource demand type labels) will lead to a decrease in the accuracy of cluster center selection, in large-scale data, the same multi-label user data needs to be recombined to form a new category; secondly, the k-medoids algorithm is used to complete the selection of the initial cluster center. Specifically, according to the resource demand type label, the users with the same label are classified into a new category to confirm the value of the number of user clusters (k). For the newly formed computing network user cluster category, the idea of ​​the k-medoids algorithm is used. Firstly, for the confirmed k clusters, the center point in the cluster is confirmed, that is, the sample closest to the center of the class average value in each cluster is taken as the center point; secondly, the average difference in the cluster is calculated so that the sum of the distances from all other points in the cluster to the sample point is minimized; finally, if the cluster center does not change or the number of iterations reaches the upper limit, the algorithm stops, and the cluster centers corresponding to the multiple clusters at the time of stopping are output as multiple initial cluster centers, otherwise the initial center in the cluster is updated, and the average difference in the cluster is calculated again.

[0066] Step S106, clustering the multiple accounts based on the multiple initial cluster centers to obtain a clustering result;

[0067] In step S106, the resource requirement type is used as a label, and the label information is used to guide the clustering process, so as to effectively avoid common problems in traditional clustering algorithms, such as local optimum and randomness of initial center point selection, and improve the accuracy and reliability of clustering results.

[0068] Optionally, based on multiple initial clustering centers, multiple accounts may be clustered using resource demand types corresponding to the multiple accounts as labels to obtain clustering results.

[0069] In an optional embodiment, multiple accounts are clustered based on multiple initial clustering centers to obtain clustering results, including: based on the multiple initial clustering centers, multiple initial populations are constructed with resource demand types corresponding to the multiple accounts as labels and the multiple accounts as particles included in the initial population; the positions of the particles included in the multiple initial populations are updated multiple times until a predetermined termination condition is reached; and the clustering results are determined based on the optimized population obtained when the predetermined termination condition is reached.

[0070] Optionally, multiple accounts are used as particles in the initial population, and the resource requirement type corresponding to each particle is used as a label, and these accounts are clustered using a clustering algorithm, so as to achieve grouping of the correlation and similarity between different accounts. Through multiple position updates, the particles will continuously adjust their positions, and finally obtain an optimized population when the predetermined termination condition is reached, thereby determining the final clustering result. The above method can effectively classify accounts so that accounts of the same category have similar resource requirement types, which is conducive to more refined management and personalized services of accounts. In addition, by clustering accounts, the relationship and characteristics between accounts can be better understood, so as to better understand and meet user needs.

[0071] In an optional embodiment, the positions of particles included in the multiple initial populations are updated multiple times until a predetermined termination condition is reached, including: based on the resource demand types corresponding to the multiple accounts, a sparrow algorithm is used to perform a first position update on the particles included in the multiple initial populations to obtain multiple first optimized populations; an elite reverse learning algorithm is used to perform a second position update on the particles included in the multiple first optimized populations to obtain multiple second optimized populations; based on the multiple second optimized populations, a K-means clustering algorithm is used to cluster the multiple accounts to obtain multiple third optimized populations; the multiple third optimized populations are used as new multiple initial populations, and the above operations are repeated until the predetermined termination condition is reached.

[0072] Optionally, treating accounts as particles and assigning them to different initial populations based on resource demand types helps to explore optimal solutions in different environments and increase the global search capability of the algorithm. By imitating the predation and vigilance behaviors of sparrows, the positions of population particles are updated, aiming to improve the distribution range of population particles through group collaboration and competition, avoid local optimal traps, and enhance the global convergence of the algorithm. By adopting the elite reverse learning algorithm, the historical optimal solution can be introduced for reverse learning, further adjusting the position of population particles, increasing population diversity, and preventing the algorithm from converging to the local optimal solution too early. After optimization by the sparrow algorithm and the elite reverse learning algorithm, the accounts are further clustered using the K-means clustering algorithm to refine the clustering results and improve the accuracy of clustering. The optimized population is used as the new initial population, and the above process is repeated until the predetermined termination condition is met, such as reaching the maximum number of iterations or the stability of the clustering results. When the termination condition is met, the final third optimized population will reflect the most optimized clustering result, that is, the best clustering state of the similarity of resource demand types between accounts. In an optional embodiment, based on multiple second optimized populations, a K-means clustering algorithm is used to cluster multiple accounts to obtain multiple third optimized populations, including: using an adaptive t-distribution method to adjust the positions of cluster centers in multiple second optimized populations to obtain multiple fourth optimized populations; based on multiple fourth optimized populations, a K-means clustering algorithm is used to cluster multiple accounts to obtain multiple third optimized populations.

[0073] In the above method, the position of the cluster center in the K-means clustering algorithm can be further optimized by introducing the adaptive t distribution method. By adjusting the cluster center through the adaptive t distribution, the accuracy and stability of the K-means clustering algorithm can be significantly improved. This method of dynamically adjusting the position of the cluster center can help the algorithm better explore the data space during the iteration process, avoid premature convergence to the local optimal solution, and thus obtain a more reasonable and accurate clustering result. The combination of the improved sparrow search algorithm and the adaptive t distribution enables the entire clustering process to not only process large-scale data, but also effectively deal with noise and anomalies in the data, thereby enhancing the robustness of the algorithm. This means that the algorithm can still maintain a high clustering performance when facing imperfect or complex data. For the clustering of computing network users, this optimized clustering method can more accurately identify the differences in user needs and provide a more accurate basis for the optimal allocation of resources.

[0074] Optionally, in the clustering of computing network users, according to the confirmed initial clustering centers, the traditional k-means clustering method is improved based on the sparrow algorithm, and the elite reverse learning is used to optimize the initialization population to increase the population diversity; through the adaptive t distribution, it is avoided to fall into the local optimum, and finally the computing network user clustering is completed.

[0075] Specifically, the population is initialized first, including setting the maximum number of iterations MaxCycle, the population size C, the number of discoverers DN, the number of vigilants PN, and the alarm threshold PT.

[0076] Further calculate the fitness value. According to the confirmed initial cluster center, calculate the fitness value of each sparrow, find the optimal fitness value and the worst fitness value and the corresponding position, where in the sparrow algorithm, the sparrow is the particle in the population.

[0077] The sparrow search algorithm achieves parameter optimization through sparrow predation and anti-predation operations. The initial position of the sparrow is expressed as follows:

[0078]

[0079] Where n is the number of sparrows, that is, the number of users that need to be classified; d is the dimension of the variable, that is, the number of resource demand types. The position of the sparrow is X = [x i1 ,x i2 ,...,x in ], the fitness value is calculated as follows.

[0080] G X =[g(X1)...g(X n )] T (5)

[0081] Among them, g(X n ) is the fitness value of the nth sparrow, G X Represents the set consisting of the fitness values ​​corresponding to all sparrows.

[0082] Further update the position of the sparrows. From the sparrows with better fitness values, select DN sparrows as discoverers, the remaining sparrows as joiners, and select PN as scouts to update their positions respectively. Calculate the fitness value and update the position of the sparrows through elite reverse learning, aiming to increase the diversity of the population.

[0083] Among them, the function and position update calculation formula of the discoverer, joiner, and scout of the sparrow algorithm are as follows:

[0084] The finder completes the search for food, and the calculation method corresponding to the finder's behavior is as follows:

[0085]

[0086] in, represents the updated position of the i-th discoverer in the j-dimension; The current position of the i-th discoverer in the j-dimension; MaxCycle is the maximum number of iterations, war∈[0,1] is the warning value, PT∈[0.5,1] is the safety value, α∈(0,1] is a random number, Q is a random number that obeys the normal distribution, and L is a 1×d matrix with all elements set to 1.

[0087] Joiners follow discoverers to find food, and the calculation method corresponding to the follower behavior is as follows:

[0088]

[0089] in, represents the updated position of the i-th joiner in the j-dimension; The current position of the i-th joiner in the j-dimension; is the optimal position of the t+1th discoverer, is the global worst position at the tth iteration, A represents a 1×d matrix whose elements are all 1 or -1, and A + =A T (AA T ) -1 .

[0090] The scout determines the anti-predator behavior. The calculation method for the scout behavior is as follows:

[0091]

[0092] in, represents the updated position of the i-th scout in the j dimension; The current position of the i-th scout in the j dimension; is the global optimal value, β is the step size control parameter and is a normally distributed random number with a mean of 0 and a variance of 1, γ∈[-1,1], g i ,g b ,g w They represent the current fitness value, the optimal fitness value, and the worst fitness value respectively.

[0093] The elite reverse learning calculation formula is as follows:

[0094] Let x=(x1,x2,...,x d ) is an elite individual in d-dimensional space, then the reverse point Defined as:

[0095]

[0096] In the formula, g is a random number between (0,1), is the boundary of the search space, and the reverse solution is processed out of bounds. The calculation formula is as follows:

[0097]

[0098] Among them, rand(lb j ,ub j ) means in lb j To ub j Pick a random number within the range of .

[0099] Further t-distribution mutation is performed. According to the rand value obtained by elite reverse learning, if rand < p, adaptive t-distribution mutation is performed to prevent the algorithm from falling into a local optimal solution. Otherwise, go directly to the next step. Among them, the value of p can be 0.05 based on historical experience.

[0100] Adaptive t-distribution mutation is performed on the sparrow positions to balance the algorithm's global exploration and local development capabilities. The adaptive t-distribution calculation formula is as follows.

[0101]

[0102] Here, t(iter) is the t distribution with iter as the parameter degree of freedom.

[0103] Further iterative judgment is performed to determine whether the iteration is completed. If not, return to step 2. If it is completed, the network energy user clustering is completed.

[0104] Step S108, determining classification results of the multiple accounts based on the clustering results.

[0105] In step S108, the classification result of the account is determined through the clustering result, which can help to better understand the user group of the computing network, thereby providing more personalized services, optimizing resource allocation, and improving resource utilization efficiency.

[0106] Through the above steps S102 to S108, it is possible to determine multiple initial clustering centers based on the resource demand type of each energy demand account, and cluster the accounts based on the determined initial clustering centers to obtain clustering results that are more in line with the user's resource needs. This allows a better understanding of the demand characteristics of each account by determining the resource demand type, provides a clear direction for subsequent clustering and classification, and thereby improves the technical effect of user classification accuracy, thereby solving the technical problem of low accuracy of classification results in related technologies that use clustering to classify energy demand accounts based on randomly selected initial clustering centers.

[0107] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode: Figure 3 is a flowchart of an optional account processing method based on network energy demand according to an embodiment of the present invention, such as Figure 3 As shown, the method includes:

[0108] S1: Mapping computing network energy user requirements to determine the resource requirements of each computing network energy user. From the user side, combined with computing network energy resource indicators and task applications, the computing network energy user requirements mapping is defined, and the computing network energy user requirements are mapped into high-sensitivity green computing requirements, low-sensitivity green computing requirements, high-sensitivity green network requirements, and low-sensitivity green network requirements. Figure 2 shown.

[0109] S2: The computing network can label user indicators, determine the resource demand type of each user, and use the resource demand type as the user label. Figure 4 is an optional flow chart of tagging network energy user indicators according to an embodiment of the present invention, such as Figure 4 As shown, the network energy index matching degree of all account data is calculated with four clearly defined resource demand types and sorted; secondly, by comparing the matching degree change of each account relative to the four resource demand types with the threshold setting method, one or more labels are added to the user data to complete the determination of the user demand type label, wherein the specific calculation method of the matching degree and the matching degree change is the same as mentioned above and will not be repeated here.

[0110] Specifically, by setting the deviation threshold α, the resource demand type of each user is determined, and the resource demand type is used as the label of the corresponding user to achieve labeling of user data. The judgment rule for each user is as follows: If the matching degree change of any account relative to any resource demand type is cha i ≤α, add user data tag type, for those that meet cha i All user resource demand types within the range of ≤α are defined as user labels (i.e. resource demand type labels). Otherwise, the user demand type with the best matching degree is defined as the only user label. Finally, the user index labeling of the computing network is completed, providing a basis for the selection of the initial clustering center.

[0111] S3: Initial cluster center selection. After completing the labeling of user data, reorganize the user categories and use the k-medoids algorithm to cluster multi-label user data to achieve initial center selection for large-scale data clustering.

[0112] In the selection of the initial cluster center, firstly, considering that multi-label user data (i.e., a computing network user may correspond to multiple resource demand type labels) will lead to a decrease in the accuracy of cluster center selection, in large-scale data, the same multi-label user data needs to be recombined to form new categories; secondly, the k-medoids algorithm is used to complete the selection of the initial cluster center. Figure 5 This is an optional initial cluster center screening flow chart according to an embodiment of the present invention, specifically:

[0113] S31: User category reorganization: According to the resource demand type label, users with the same label are classified into a new category to confirm the value of the number of user clusters (k).

[0114] S32: Confirmation of initial cluster center. For the newly formed computing network user clustering category, the idea of ​​k-medoids algorithm is used. First, for the confirmed k clusters, confirm the center point within the cluster, that is, the sample in each cluster that is closest to the center of the class average value is taken as the center point; secondly, calculate the average difference within the cluster so that the sum of the distances from all other points in the cluster to the sample point is minimized; finally, if the cluster center does not change or the number of iterations reaches the upper limit, the algorithm stops, otherwise update the initial center within the cluster and return to calculate the average difference within the cluster.

[0115] S4: Clustering of network users. The cluster center of step S3 is used as the initial center, and the k-means clustering algorithm is optimized using the improved sparrow algorithm. Specifically, elite reverse learning is performed on the sparrow population to enhance population diversity, and adaptive t distribution is used to avoid falling into local optimality, so as to complete accurate clustering of network users. Figure 6 is an optional network user clustering flow chart according to an embodiment of the present invention, such as Figure 6 As shown, specifically including:

[0116] S41: Population initialization, including setting the maximum number of iterations MaxCycle, population size C, number of discoverers DN, number of vigilants PN, and alarm threshold PT.

[0117] S42: Calculate the fitness value. According to the confirmed initial cluster center, calculate the fitness value of each sparrow, and find the best fitness value and the worst fitness value and the corresponding position.

[0118] Step S43: Position update. From the sparrows with better fitness values, select DN sparrows as discoverers, the remaining sparrows as joiners, and select PN as scouts to update their positions respectively. Calculate the fitness value and update the sparrow positions through elite reverse learning to increase population diversity.

[0119] Step S44: t distribution mutation. According to the rand value obtained by elite reverse learning, if rand < p, adaptive t distribution mutation is performed to prevent the algorithm from falling into the local optimal solution, and return to step S42. Otherwise, go directly to the next step. Among them, the value of p can be 0.05 based on historical experience.

[0120] Step S45: Iteration determination. Determine whether the iteration is finished. If not, return to step 2. If finished, the network energy user clustering is completed.

[0121] In this embodiment, a clustering method based on multi-label information on the user side of the computing network energy is designed in view of the fact that the computing network energy user clustering methods in the related arts mostly start from the supply side, ignore the mapping relationship between task applications and user needs, and have deficiencies in both initial center selection and local optimal control. First, combining computing power, network and energy indicators, from the user side, the user demand types of computing network energy are defined according to task applications, namely, high-sensitivity green computing demand, low-sensitivity green computing demand, high-sensitivity green network demand, and low-sensitivity green network demand, in order to solve the problem of vague computing network energy user demand; secondly, the matching degree of all user data and the four types of user demand is calculated, and the user data is labeled by threshold setting to provide a basis for the selection of initial clustering centers; thirdly, multi-label data is clustered, and the k-medoids algorithm is used to divide all multi-label vectors into k clusters to complete the rational selection of initial centers for large-scale data; finally, the existing centers are used as the initial centers for clustering, and the k-means clustering algorithm is optimized using the improved sparrow algorithm. The elite reverse learning is performed through the sparrow population to enhance the population diversity. The adaptive t distribution is used to avoid falling into the local optimum, and the precise clustering of computing network energy users is completed to achieve efficient clustering convergence.

[0122] This embodiment can achieve at least one of the following effects: 1) From the user side, consider computing network resources and energy resources at the same time to meet the needs of accurate user demand mapping; use multi-label information to optimize the initial cluster center selection and solve the problem of reasonable initial center selection; optimize the k-means clustering algorithm by improving the sparrow algorithm to enhance population diversity, avoid falling into the local optimal solution, and achieve efficient clustering convergence. 2) From the user side, fully understand the relationship between "task-user demand-resource", build a three-dimensional computing network energy user demand mapping system including task layer, user demand layer, and indicator layer, divide computing network energy task types, and clearly define the user demand layer, including high-sensitivity green computing demand, low-sensitivity green computing demand, high-sensitivity green network demand, and low-sensitivity green network demand, to achieve accurate mapping of computing network energy user demand. 3) For the constructed computing network energy user demand mapping system, the computing network energy indicator matching calculation method is used to match all user data with four clearly defined user demand types; through the method of result sorting, demand type matching value comparison, and threshold setting, one or more labels are added to the user data to reduce the scale of user data. 4) User data is reorganized through label information to form new categories. On this basis, the k-medoids algorithm is used to complete the selection of initial clustering centers to ensure the rationality of the initial center selection. 5) An improved k-means computing network energy user clustering method based on the sparrow algorithm is designed. The elite reverse learning is used to optimize the initialization population and increase the population diversity; the adaptive t distribution is used to avoid falling into the local optimum and improve the efficient and accurate clustering of computing network energy users.

[0123] In this embodiment, an account processing device based on network energy demand is also provided, which is used to implement the above embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the terms "module" and "device" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0124] According to an embodiment of the present invention, there is also provided an apparatus embodiment for implementing the above-mentioned account processing method based on computing network energy demand. Figure 7 is a structural diagram of an account processing device based on network energy demand according to an embodiment of the present invention, such as Figure 7 As shown, the above-mentioned account processing device based on computing network energy demand includes: a type determination module 700, a cluster center determination module 702, an account clustering module 704, and an account classification module 706, wherein:

[0125] The type determination module 700 is used to determine the resource demand types corresponding to the multiple accounts with computing network energy demand, wherein the computing network energy demand is used to indicate that the corresponding account has computing power resources, network resources, and energy resources;

[0126] A cluster center determination module 702, connected to the type determination module 700, is used to determine a plurality of initial cluster centers for clustering the plurality of accounts based on the resource demand types respectively corresponding to the plurality of accounts;

[0127] The account clustering module 704 is connected to the cluster center determination module 702 and is used to cluster multiple accounts based on multiple initial cluster centers to obtain clustering results;

[0128] The account classification module 706 is connected to the account clustering module 704 and is used to determine classification results of multiple accounts based on the clustering results.

[0129] In the embodiment of the present invention, a type determination module 700 is provided to determine the resource demand types corresponding to the multiple accounts with computing network energy demand, wherein the computing network energy demand is used to indicate that the corresponding account has computing power resources, network resources, and energy resources; a clustering center determination module 702 is connected to the type determination module 700, and is used to determine multiple initial clustering centers for clustering the multiple accounts based on the resource demand types corresponding to the multiple accounts; an account clustering module 704 is connected to the clustering center determination module 702, and is used to cluster the multiple accounts based on the multiple initial clustering centers to obtain clustering results; an account classification module 706 is connected to the account clustering module 704, and is used to determine the classification results of the multiple accounts based on the clustering results, thereby achieving the purpose of determining multiple initial clustering centers based on the resource demand type of each energy demand account, and clustering the accounts based on the determined initial clustering centers to obtain clustering results that are more in line with the user's resource demand, thereby achieving the technical effect of improving the accuracy of the account clustering results, and further solving the technical problem of low accuracy of the classification results in the related art based on the randomly selected initial clustering centers and the classification of energy demand accounts by clustering.

[0130] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0131] It should be noted that the above-mentioned type determination module 700, cluster center determination module 702, account clustering module 704, and account classification module 706 correspond to steps S102 to S108 in the embodiment, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment. It should be noted that the above-mentioned modules can be run in a computer terminal as part of the device.

[0132] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0133] The above-mentioned account processing device based on computing network energy demand may also include a processor and a memory. The above-mentioned type determination module 700, cluster center determination module 702, account clustering module 704, account classification module 706, etc. are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to realize the corresponding functions.

[0134] The processor includes a kernel, which retrieves the corresponding program module from the memory. The kernel may be one or more. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0135] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute any of the above-mentioned account processing methods based on computing network energy demand.

[0136] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group, and the non-volatile storage medium includes a stored program.

[0137] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: determine the resource requirement types corresponding to multiple accounts with computing network energy requirements, wherein the computing network energy requirements are used to indicate that the corresponding accounts have computing resources, network resources, and energy resources; based on the resource requirement types corresponding to the multiple accounts, determine multiple initial clustering centers for clustering the multiple accounts; based on the multiple initial clustering centers, cluster the multiple accounts to obtain clustering results; based on the clustering results, determine the classification results of the multiple accounts.

[0138] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein when the program is run, any of the above-mentioned account processing methods based on computing network energy demand is executed.

[0139] According to an embodiment of the present application, an embodiment of a computer program product is also provided. Optionally, in this embodiment, the computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements any of the steps of the above-mentioned account processing method based on computing network energy demand.

[0140] Optionally, the above-mentioned computer program product, when executed on a data processing device, is suitable for executing a program initialized with the following method steps: determining the resource demand types corresponding to multiple accounts with computing network energy requirements, wherein the computing network energy requirements are used to indicate that the corresponding accounts have computing power resources, network resources, and energy resources; based on the resource demand types corresponding to the multiple accounts, determining multiple initial clustering centers for clustering the multiple accounts; based on the multiple initial clustering centers, clustering the multiple accounts to obtain clustering results; based on the clustering results, determining the classification results of the multiple accounts.

[0141] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: determining resource demand types corresponding to multiple accounts with computing network energy requirements, wherein the computing network energy requirements are used to indicate that the corresponding accounts have computing power resources, network resources, and energy resources; based on the resource demand types corresponding to the multiple accounts, determining multiple initial clustering centers for clustering the multiple accounts; based on the multiple initial clustering centers, clustering the multiple accounts to obtain clustering results; based on the clustering results, determining classification results for the multiple accounts.

[0142] The above sequence of the embodiments of the present invention is for description only and does not represent the superiority or inferiority of the embodiments.

[0143] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0144] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the above modules can be a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0145] The modules described above as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0146] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software functional modules.

[0147] If the above-mentioned integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0148] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An account processing method based on network energy demand, characterized in that: include: Determine resource demand types corresponding to multiple accounts with computing network energy requirements, wherein the computing network energy requirements are used to indicate that the corresponding accounts have computing power resources, network resources, and energy resources; Determining a plurality of initial clustering centers for clustering the plurality of accounts based on the resource demand types respectively corresponding to the plurality of accounts; Based on the multiple initial cluster centers, clustering the multiple accounts to obtain a clustering result; Based on the clustering result, classification results of the multiple accounts are determined.

2. The method according to claim 1, characterized in that: The determining of resource demand types corresponding to the multiple accounts having network computing energy requirements includes: Obtain resource indicators corresponding to the multiple accounts respectively, wherein the resource indicators include computing power indicators corresponding to the computing power resources, network indicators corresponding to the network resources, and energy indicators corresponding to the energy resources; Based on the resource indicators respectively corresponding to the multiple accounts, the resource demand types respectively corresponding to the multiple accounts are determined.

3. The method according to claim 2, characterized in that The determining, based on the resource indicators respectively corresponding to the multiple accounts, the resource demand types respectively corresponding to the multiple accounts comprises: The resource requirement type of any account among the multiple accounts is obtained in the following manner: Based on the resource indicators of any of the accounts, determine an indicator score of any of the accounts, wherein the indicator score includes a computing power indicator score, a network indicator score, and an energy indicator score; Based on the preset indicator scores corresponding to the multiple resource demand types and the indicator score of any one of the accounts, determining the matching degree between any one of the accounts and the multiple resource demand types, wherein the preset indicator score includes a preset computing power indicator score, a preset network indicator score, and a preset energy indicator score; Determining the resource demand type of any one of the accounts based on the matching degree between any one of the accounts and the multiple resource demand types respectively; The resource demand types respectively corresponding to the multiple accounts are obtained by obtaining the resource demand type of any one of the accounts.

4. The method according to claim 3, characterized in that: In the case where there are multiple computing power indicators, multiple network indicators, and multiple energy indicators, determining the indicator score of any account based on the resource indicators of any account includes: Determine weight values ​​corresponding to a plurality of computing power indicators respectively; perform weighted calculation on the plurality of computing power indicators and the weight values ​​corresponding to the plurality of computing power indicators respectively to obtain the computing power indicator score; Determine weight values ​​corresponding to a plurality of network indicators respectively; perform weighted calculation on the plurality of network indicators and the weight values ​​corresponding to the plurality of network indicators respectively to obtain the network indicator score; Determine the weight values ​​corresponding to the multiple energy indicators respectively; perform weighted calculation on the multiple energy indicators and the weight values ​​of the multiple energy indicators respectively to obtain the energy indicator score.

5. The method according to claim 3, characterized in that: The determining the resource requirement type of any one of the accounts based on the matching degree between the any one of the accounts and the multiple resource requirement types respectively includes: The matching degree change of any one of the multiple resource demand types for any one of the account is obtained by: Determine a maximum value of the matching degrees between the multiple accounts and any resource demand type as the optimal matching degree; Based on the matching degree between any one of the accounts and any one of the resource demand types, and the optimal matching degree, obtaining a matching degree change amount of any one of the accounts relative to any one of the resource demand types, wherein the matching degree change amount is used to indicate a degree of change of the matching degree between the corresponding account and the resource demand type relative to the corresponding optimal matching degree; By obtaining a change in the degree of matching of any one of the accounts relative to any one of the resource demand types, the change in the degree of matching of the resource indicator of any one of the accounts relative to the multiple resource demand types is obtained; The resource demand type of any one of the accounts is determined based on the changes in the matching degrees of the resource indicators of any one of the accounts relative to the multiple resource demand types.

6. The method according to claim 1, characterized in that The determining a plurality of initial clustering centers for clustering the plurality of accounts based on the resource demand types respectively corresponding to the plurality of accounts comprises: The resource demand types respectively corresponding to the multiple accounts are used as labels, and the multiple accounts are clustered using a K-center point clustering algorithm to obtain the multiple initial cluster centers.

7. The method according to claim 1, characterized in that The clustering of the multiple accounts based on the multiple initial cluster centers to obtain a clustering result includes: Based on the multiple initial cluster centers, using the resource demand types respectively corresponding to the multiple accounts as labels and using the multiple accounts as particles included in the initial population, constructing multiple initial populations; Performing multiple position updates on particles included in the multiple initial populations until a predetermined termination condition is reached; The clustering result is determined based on the optimized population obtained when the predetermined termination condition is reached.

8. The method according to claim 7, characterized in that The step of updating the positions of the particles included in the multiple initial populations for multiple times until a predetermined termination condition is reached includes: Based on the resource demand types respectively corresponding to the multiple accounts, using a sparrow algorithm to perform a first position update on the particles included in the multiple initial populations to obtain multiple first optimized populations; Using an elite reverse learning algorithm to perform a second position update on the particles included in the plurality of first optimized populations to obtain a plurality of second optimized populations; Based on the multiple second optimized populations, clustering the multiple accounts using a K-means clustering algorithm to obtain multiple third optimized populations; The plurality of third optimized populations are used as a plurality of new initial populations, and the above operations are repeatedly performed until the predetermined termination condition is reached.

9. The method according to claim 8, characterized in that The method of clustering the multiple accounts based on the multiple second optimized populations using a K-means clustering algorithm to obtain multiple third optimized populations includes: Adopting an adaptive t distribution method to adjust the positions of cluster centers in the plurality of second optimized populations to obtain a plurality of fourth optimized populations; Based on the multiple fourth optimized populations, the multiple accounts are clustered using the K-means clustering algorithm to obtain the multiple third optimized populations.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, which are suitable for being loaded by a processor and executing the account processing method based on computing network energy demand as described in any one of claims 1 to 9.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the account processing method based on computing network energy demand as described in any one of claims 1 to 9 are implemented.

Images