Method and system for private cloud resource demand prediction and configuration based on spatiotemporal sequences

Abstract

The application provides a private cloud resource demand prediction and configuration method and system based on a space-time sequence, obtains cloud resource configuration history data of all private cloud management domains from cloud resource configuration history logs of the respective private cloud management domains, pre-processes and periodically changes the resource configuration history data, extracts corresponding prediction sample data, and provides sufficient data for subsequent training of a space-time sequence model; based on the trained space-time sequence model, cloud resource configuration demand data of all private cloud management domains in a future preset time period is predicted, so as to determine cloud resource configuration feature information corresponding to each private cloud management domain, and all private cloud management domains with the same cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types, so as to form a corresponding management domain cluster, and the historical data and data features of the management domain cluster are used to predict the cloud resources of the management domains belonging to different management domain clusters with strong features; and based on the future cloud resource demand information of each private cloud management domain, a comparison analysis is performed with a configuration utilization rate threshold, all cloud resources under the private cloud management domain are optimized and adjusted, and the utilization efficiency of the cloud resources is improved.

Description

Technical Field

[0001] The present invention relates to the field of cloud resources, and in particular to a method and system for predicting and configuring private cloud resource demand based on spatiotemporal sequences. Background Art

[0002] As cloud service providers continue to expand their cloud computing resources, scientifically and effectively forecasting cloud resource demand is crucial for better planning and management of cloud computing resources. Currently, cloud resource demand allocation is primarily achieved through traditional machine learning or deep learning models based on time series. However, these approaches focus on trends and periodicity in the temporal dimension while ignoring spatial and temporal correlations. They also fail to address spatial heterogeneity and model regional differences. This results in a decrease in the generalization ability of the forecasting model, making it impossible to accurately and timely provide cloud resource allocation for different users, reducing cloud resource utilization efficiency and increasing the probability of idle cloud resources. Summary of the Invention

[0003] The present invention aims to provide a method and system for predicting and configuring private cloud resource requirements based on spatiotemporal sequences. The method obtains historical cloud resource configuration data for all private cloud management domains from their respective cloud resource configuration history logs, preprocesses the historical resource configuration data and identifies periodic changes, extracts corresponding prediction sample data, and provides sufficient data for subsequent training of a spatiotemporal sequence model. Based on the trained spatiotemporal sequence model, the method predicts the cloud resource configuration requirements for all private cloud management domains within a preset future time period. Cloud resource configuration feature information corresponding to each private cloud management domain is thereby determined. Furthermore, all private cloud management domains with similar cloud resource configuration feature information are grouped from all private cloud management domains based on different regions and business types to form corresponding management domain clusters. The method utilizes the historical data and data features of the management domain clusters to perform highly characteristic cloud resource predictions for management domains belonging to different management domain clusters. Furthermore, based on the future cloud resource demand information of each private cloud management domain, a comparative analysis is performed against a configuration utilization threshold to optimize and adjust all cloud resources within the private cloud management domain, thereby improving cloud resource utilization efficiency.

[0004] The present invention is achieved through the following technical solutions:

[0005] The private cloud resource demand prediction and configuration method based on spatiotemporal series includes:

[0006] Based on the cloud resource configuration history logs of each of the several private cloud management domains, the cloud resource configuration history data of each of the private cloud management domains is obtained, and the cloud resource configuration history data is preprocessed; based on the preprocessed cloud resource configuration history data, periodic change identification processing is performed to extract corresponding prediction sample data;

[0007] Based on the predicted sample data, a spatiotemporal sequence model is trained, and based on the trained spatiotemporal sequence model, cloud resource configuration demand data for all private cloud management domains within a preset future time period is predicted; based on the cloud resource configuration demand data, cloud resource configuration feature information corresponding to each private cloud management domain is determined; based on the cloud resource configuration feature information corresponding to all private cloud management domains, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters;

[0008] Based on the future cloud resource demand information of each private cloud management domain, a comparison analysis is performed with the configured utilization rate threshold, and all cloud resources under the private cloud management domain are optimized and adjusted.

[0009] Optionally, based on the cloud resource configuration history logs of each of the several private cloud management domains, obtaining the cloud resource configuration history data of each of the private cloud management domains, and preprocessing the cloud resource configuration history data; performing periodic change identification processing on the preprocessed cloud resource configuration history data to extract corresponding prediction sample data, including:

[0010] Analyze the cloud resource configuration history logs of each of several private cloud management domains to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval; perform cloud resource configuration process analysis on all cloud resource configuration events to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time when the configuration event occurred, as the cloud resource configuration history data; and perform interference noise data cleaning preprocessing and missing data filling preprocessing on the cloud resource configuration history data.

[0011] The cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the cloud resource configuration historical data that has completed all preprocessing is subjected to bad point data elimination processing to obtain corresponding prediction sample data.

[0012] Optionally, the cloud resource configuration historical data is sequentially preprocessed to fill in missing data, including:

[0013] Extract two valid data adjacent to and before the data position corresponding to the missing data as the first valid data and the second valid data;

[0014] Acquire a data difference between the first valid data and the second valid data according to the first valid data and the second valid data;

[0015] When the data difference between the first valid data and the second valid data is 0, the first valid data and the second valid data are weighted averaged to obtain a weighted average, and the weighted average is used as the first filling data to fill the missing data;

[0016] When the data difference between the first valid data and the second valid data is not 0, the data difference between the first valid data and the second valid data is used as a difference threshold;

[0017] Retrieving, from the historical data, historical data whose difference with the first valid data exceeds the difference threshold as the first historical data;

[0018] Retrieving, from the historical data, historical data whose difference with the second valid data exceeds the difference threshold as the second historical data;

[0019] The second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula:

[0020] Wherein, X represents the second filling data; X p represents the average value of valid data of historical data of cloud resource configuration; n represents the number of data of the first historical data; X 01i represents the data value corresponding to the first historical data of the i-th item; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02p represents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0021] Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data;

[0022] The missing data is filled using the second filling data.

[0023] Optionally, based on the predicted sample data, a spatiotemporal sequence model is trained, and based on the trained spatiotemporal sequence model, cloud resource configuration demand data for all private cloud management domains in a future preset time period is predicted; based on the cloud resource configuration demand data, cloud resource configuration feature information corresponding to each private cloud management domain is determined; based on the cloud resource configuration feature information corresponding to all private cloud management domains, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types, thereby forming corresponding management domain clusters, including:

[0024] Inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model, and predicting cloud resource configuration demand data for all private cloud management domains within a future preset time period based on the trained Graph WaveNet model; wherein the cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain for each unit time interval within the future preset time period;

[0025] Analyzing the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period;

[0026] Based on the demand period distribution characteristic information corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristic information, region and business type are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters.

[0027] Optionally, based on the future cloud resource demand information of each private cloud management domain, a comparison analysis is performed with a configured utilization threshold, and all cloud resources under the private cloud management domain are optimized and adjusted, including:

[0028] The future cloud resource demand information of each private cloud management domain and the corresponding configuration utilization rate thresholds of all types of cloud resources are compared and analyzed to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time intervals.

[0029] A private cloud resource demand prediction and configuration system based on spatiotemporal sequences, including:

[0030] a historical data acquisition and preprocessing module configured to acquire cloud resource configuration historical data of each of the private cloud management domains based on cloud resource configuration historical logs of the private cloud management domains, and to preprocess the cloud resource configuration historical data;

[0031] a sample data generation module configured to perform periodic change identification processing based on the cloud resource configuration historical data after the preprocessing, and to extract corresponding prediction sample data;

[0032] a cloud resource configuration demand prediction module configured to train a spatiotemporal sequence model based on the prediction sample data, and to predict cloud resource configuration demand data of each of the private cloud management domains in a future preset time period based on the trained spatiotemporal sequence model;

[0033] a cloud resource configuration feature information determination module configured to determine cloud resource configuration feature information corresponding to each of the private cloud management domains based on the cloud resource configuration demand data;

[0034] a management domain cluster generation module configured to divide all private cloud management domains with the same cloud resource configuration feature information from all the private cloud management domains according to different regions and business types based on the cloud resource configuration feature information corresponding to each of the private cloud management domains, to thereby form corresponding management domain clusters;

[0035] a cloud resource configuration adjustment module configured to compare and analyze future cloud resource demand information of each of the private cloud management domains with a configuration utilization rate threshold, and to optimize and adjust all cloud resources under the private cloud management domains.

[0036] Optionally, the historical data acquisition and preprocessing module is configured to acquire cloud resource configuration historical data of each of the private cloud management domains based on cloud resource configuration historical logs of the private cloud management domains, and to preprocess the cloud resource configuration historical data, including:

[0037] analyzing the cloud resource configuration historical logs of each of the private cloud management domains to obtain all cloud resource configuration events of each of the private cloud management domains occurring in a preset historical time interval, performing cloud resource configuration process analysis on all the cloud resource configuration events respectively to obtain cloud resource types and quantities corresponding to each of the cloud resource configuration events and configuration event occurrence times, taking the cloud resource configuration historical data as the cloud resource types, the quantities, and the configuration event occurrence times, and sequentially performing interference noise data cleaning preprocessing and missing data filling preprocessing on the cloud resource configuration historical data;

[0038] the sample data generation module is configured to perform periodic change identification processing based on the cloud resource configuration historical data after the preprocessing, and to extract corresponding prediction sample data, including:

[0039] The cloud resource configuration historical data after all preprocessing is processed three times of exponential smoothing to obtain periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the cloud resource configuration historical data after all preprocessing is processed for bad data elimination to obtain corresponding prediction sample data.

[0040] Optionally, the cloud resource configuration historical data is sequentially subjected to missing data filling preprocessing, comprising:

[0041] extracting two adjacent effective data before and after the data position corresponding to the missing data as first effective data and second effective data;

[0042] obtaining a data difference value between the first effective data and the second effective data according to the first effective data and the second effective data;

[0043] when the data difference value between the first effective data and the second effective data is 0, the first effective data and the second effective data are subjected to weighted average processing to obtain a weighted average value, and the weighted average value is used as first filling data to fill the missing data;

[0044] when the data difference value between the first effective data and the second effective data is not 0, the data difference value between the first effective data and the second effective data is used as a difference threshold value;

[0045] in the historical data, historical data whose difference value with the first effective data exceeds the difference threshold value is called as first historical data;

[0046] in the historical data, historical data whose difference value with the second effective data exceeds the difference threshold value is called as second historical data;

[0047] second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula:

[0048] wherein, X represents the second filling data; X p represents the average value of the effective data of the cloud resource configuration historical data; n represents the data number of the first historical data; X 01i represents the data value corresponding to the i-th first historical data; X 01 represents the data value corresponding to the first effective data; X 02 represents the data value corresponding to the second effective data; X 01b represents the standard deviation of the data values corresponding to the n first historical data; X 02b represents the standard deviation of the data values corresponding to the m second historical data; X 02prepresents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0049] Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data;

[0050] The missing data is filled using the second filling data.

[0051] Optionally, the cloud resource configuration demand prediction module is used to train a spatiotemporal sequence model based on the prediction sample data, and predict the cloud resource configuration demand data of all private cloud management domains within a preset future time period based on the trained spatiotemporal sequence model, including:

[0052] Inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model, and predicting cloud resource configuration demand data for all private cloud management domains within a future preset time period based on the trained Graph WaveNet model; wherein the cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain for each unit time interval within the future preset time period;

[0053] The cloud resource configuration characteristic information determination module is used to determine the cloud resource configuration characteristic information corresponding to each private cloud management domain based on the cloud resource configuration requirement data, including:

[0054] Analyzing the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period;

[0055] The management domain cluster generation module is used to divide all private cloud management domains with similar cloud resource configuration feature information from all private cloud management domains according to different regions and business types based on the cloud resource configuration feature information corresponding to each of the private cloud management domains, thereby forming corresponding management domain clusters, including:

[0056] Based on the demand period distribution characteristic information corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristic information, region and business type are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters.

[0057] Optionally, the cloud resource configuration adjustment module is configured to compare and analyze future cloud resource demand information of each private cloud management domain with a configured utilization rate threshold, and optimize and adjust all cloud resources under the private cloud management domain, including:

[0058] The future cloud resource demand information of each private cloud management domain and the corresponding configuration utilization rate thresholds of all types of cloud resources are compared and analyzed to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time intervals.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The spatiotemporal sequence-based private cloud resource demand prediction and configuration method and system provided in the present application obtains the cloud resource configuration historical data of all private cloud management domains from their respective cloud resource configuration historical logs, preprocesses the resource configuration historical data and performs periodic change identification processing on the resource configuration historical data, and extracts corresponding prediction sample data to provide sufficient data for subsequent training of the spatiotemporal sequence model. Based on the trained spatiotemporal sequence model, the cloud resource configuration required data of all private cloud management domains in a preset future time period is predicted, thereby determining the cloud resource configuration characteristic information corresponding to each private cloud management domain. According to different regions and business types, all private cloud management domains with similar cloud resource configuration characteristic information are divided from all private cloud management domains to form corresponding management domain clusters. The historical data and data characteristics of the management domain clusters are used to perform cloud resource prediction with strong characteristics for management domains belonging to different management domain clusters. Based on the future cloud resource demand information of each private cloud management domain, a comparative analysis is performed with the configuration utilization rate threshold, and all cloud resources under the private cloud management domain are optimized and adjusted to improve the utilization efficiency of cloud resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. Among them:

[0062] Figure 1 A flow chart of the private cloud resource demand prediction and configuration method based on spatiotemporal sequences provided by the present invention.

[0063] Figure 2 This is a structural diagram of the private cloud resource demand prediction and configuration system based on spatiotemporal sequences provided by the present invention. DETAILED DESCRIPTION

[0064] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some, rather than all, structures related to the present application are shown in the accompanying drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0065] As used herein, the terms "comprise," "comprising," and "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or apparatus.

[0066] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] See also Figure 1 As shown, an embodiment of the present application provides a method for predicting and configuring private cloud resource requirements based on spatiotemporal sequences. The method for predicting and configuring private cloud resource requirements based on spatiotemporal sequences includes:

[0068] Based on the cloud resource configuration history logs of several private cloud management domains, the cloud resource configuration history data of all private cloud management domains is obtained and preprocessed; based on the preprocessed cloud resource configuration history data, periodic change identification processing is performed to extract corresponding prediction sample data;

[0069] Based on the predicted sample data, a spatiotemporal sequence model is trained, and based on the trained spatiotemporal sequence model, cloud resource configuration demand data for all private cloud management domains within a preset future time period is predicted; based on the cloud resource configuration demand data, cloud resource configuration feature information corresponding to each private cloud management domain is determined; based on the cloud resource configuration feature information corresponding to all private cloud management domains, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types, thereby forming corresponding management domain clusters;

[0070] Based on the future cloud resource demand information of each private cloud management domain, a comparative analysis is performed with the configured utilization threshold to optimize and adjust all cloud resources under the private cloud management domain.

[0071] The beneficial effects of the above embodiments are as follows: the method for predicting and configuring private cloud resource demand based on spatiotemporal sequences obtains cloud resource configuration historical data of all private cloud management domains from their respective cloud resource configuration history logs, preprocesses the resource configuration historical data and identifies periodic changes, and extracts corresponding prediction sample data, providing sufficient data for subsequent training of the spatiotemporal sequence model; based on the trained spatiotemporal sequence model, predicts the cloud resource configuration requirements of all private cloud management domains within a preset future time period, thereby determining cloud resource configuration feature information corresponding to each private cloud management domain; and, based on different regions and business types, divides all private cloud management domains with similar cloud resource configuration feature information from all private cloud management domains to form corresponding management domain clusters. Utilizing the historical data and data features of the management domain clusters, a highly characteristic cloud resource prediction is performed for management domains belonging to different management domain clusters; and, based on the future cloud resource demand information of each private cloud management domain, a comparative analysis is performed with a configuration utilization threshold, and all cloud resources under the private cloud management domain are optimized and adjusted to improve the utilization efficiency of cloud resources.

[0072] In another embodiment, based on the cloud resource configuration history logs of each of several private cloud management domains, cloud resource configuration history data of each of the private cloud management domains is obtained, and the cloud resource configuration history data is preprocessed; periodic change identification processing is performed on the preprocessed cloud resource configuration history data to extract corresponding prediction sample data, including:

[0073] Analyze the cloud resource configuration history logs of several private cloud management domains to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval; perform cloud resource configuration process analysis on all cloud resource configuration events to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time when the configuration event occurred, which serves as the cloud resource configuration history data; and perform interference noise data cleaning preprocessing and missing data filling preprocessing on this cloud resource configuration history data.

[0074] The cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain the periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the bad point data of the cloud resource configuration historical data that has completed all preprocessing is eliminated to obtain the corresponding prediction sample data.

[0075] The beneficial effects of the above embodiment are that the private cloud management domain contains different types of cloud resources, which may include but are not limited to computing power cloud resources, storage cloud resources, and network bandwidth cloud resources. When a terminal performs a complex computing task, it needs to rely on the cloud resources under the private cloud management domain to undertake the corresponding computing task. In this way, the private cloud management domain will generate a corresponding cloud resource configuration history log. The cloud resource configuration history logs of several private cloud management domains are analyzed to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval. The cloud resource configuration process of all cloud resource configuration events is analyzed separately to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time of occurrence of the configuration event, which is used as the cloud resource configuration history data. This ensures that the cloud resource configuration history data fully and accurately reflects the actual cloud resource configuration status of the private cloud management domain. The cloud resource configuration history data is then subjected to interference noise data cleaning preprocessing and missing data filling preprocessing, which can improve the accuracy and credibility of the cloud resource configuration history data. Then, the cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain the periodic change trend characteristic information of the cloud resource configuration historical data. In this way, the bad point data of the cloud resource configuration historical data that has completed all preprocessing is eliminated to obtain the corresponding prediction sample data, thereby improving the quality and reliability of the prediction sample data.

[0076] In another embodiment, the cloud resource configuration history data is sequentially preprocessed to fill in missing data, including:

[0077] Extract two valid data adjacent to and before the data position corresponding to the missing data as the first valid data and the second valid data;

[0078] Acquire a data difference between the first valid data and the second valid data according to the first valid data and the second valid data;

[0079] When the data difference between the first valid data and the second valid data is 0, the first valid data and the second valid data are weighted averaged to obtain a weighted average, and the weighted average is used as the first filling data to fill the missing data;

[0080] When the data difference between the first valid data and the second valid data is not 0, the data difference between the first valid data and the second valid data is used as a difference threshold;

[0081] Retrieving, from the historical data, historical data whose difference with the first valid data exceeds the difference threshold as the first historical data;

[0082] Retrieving, from the historical data, historical data whose difference with the second valid data exceeds the difference threshold as the second historical data;

[0083] The second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula:

[0084] Wherein, X represents the second filling data; X p represents the average value of valid data of historical data of cloud resource configuration; n represents the number of data of the first historical data; X 01i represents the data value corresponding to the first historical data of the i-th item; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02p represents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0085] Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data;

[0086] The missing data is filled in by using the second filling data.

[0087] The beneficial effects of the above embodiments are that the integrity of the cloud resource configuration history data is ensured by filling in the missing data, which is crucial for subsequent data analysis and decision making. The presence of missing data can affect the accuracy of data analysis, and this technical solution can effectively solve this problem. In the technical solution, two different filling strategies are adopted: when the difference value of adjacent valid data is 0, weighted average processing is adopted, and when the difference value is not 0, more complex calculation is carried out based on historical data. This strategy makes the filling process more flexible and can choose the most appropriate filling method according to different situations. When the difference value of adjacent valid data is not 0, the technical solution derives a more accurate filling value (second filling data) by calculating the difference value of the first historical data and the second historical data with the valid data, as well as the standard deviation and the average value. This method considers the influence of multiple historical data points, thereby improving the accuracy of filling. A regulation coefficient is introduced to improve adaptability: a regulation coefficient s is introduced in the technical solution, which is calculated according to the number of data of the second historical data, the standard deviation, and the difference value of the first valid data and the second valid data. The introduction of the regulation coefficient makes the filling process more flexible and adaptable, and can better handle the missing data filling problem in different situations. The entire technical solution starts from extracting adjacent valid data, to calculating data difference, obtaining historical data, calculating filling value, etc., forming a complete data processing flow. This flow not only has clear logic, but also is easy to implement, and can efficiently process a large amount of cloud resource configuration history data.

[0088] In summary, this technical solution improves the integrity and accuracy of data, enhances the flexibility and adaptability of data processing, and optimizes the data processing flow, providing strong support for subsequent data analysis and decision making.

[0089] In another embodiment, based on the predicted sample data, a spatio-temporal sequence model is trained, and based on the trained spatio-temporal sequence model, cloud resource configuration demand data of each private cloud management domain in a future preset time period is predicted; based on the cloud resource configuration demand data, cloud resource configuration feature information corresponding to each private cloud management domain is determined; based on the cloud resource configuration feature information corresponding to each private cloud management domain, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types, to form a corresponding management domain cluster, including:

[0090] The predicted sample data is input into a Graph WaveNet model to train the Graph WaveNet model. Based on the trained Graph WaveNet model, cloud resource configuration demand data for all private cloud management domains in a future preset time period is predicted. The cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain in each unit time interval in the future preset time period.

[0091] Analyze the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period;

[0092] Based on the demand period distribution characteristic information corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristic information, region and business type are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters.

[0093] The beneficial effects of the above embodiment include inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model. Based on the trained Graph WaveNet model, cloud resource configuration demand data for all private cloud management domains within a future preset time period is predicted. This allows for a comprehensive and detailed prediction of each private cloud management domain's cloud resource demand within the future preset time period. This cloud resource configuration demand data is also analyzed to obtain characteristic information about the time period distribution of each private cloud management domain's demand for different types of cloud resources, thereby quantifying the cloud resource demand of each private cloud management domain. Based on this characteristic information about the time period distribution of each private cloud management domain's demand for different types of cloud resources, all private cloud management domains with complementary cloud resource configuration relationships are selected from all private cloud management domains to form a corresponding management domain cluster. This ensures that all private cloud management domains within the management domain cluster can uniformly configure at least one type of cloud resource across all unit time intervals within the future preset time period, providing stable and reliable cloud resource adjustment services for corresponding terminals.

[0094] In another embodiment, based on the future cloud resource demand information of each private cloud management domain and comparing it with the configured utilization threshold, all cloud resources under the private cloud management domain are optimized and adjusted, including:

[0095] The future cloud resource demand information of each private cloud management domain and the corresponding configuration utilization rate thresholds of all types of cloud resources are compared and analyzed to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time intervals.

[0096] The beneficial effect of the above embodiment is to compare and analyze the future cloud resource demand information of each private cloud management domain and the configuration utilization rate thresholds corresponding to all types of cloud resources, and obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period. In this way, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time interval, which can take into account the continuous and stable cloud resource adjustment of different management domains and the targeted guarantee that the management domain obtains sufficient cloud resource quantity configuration.

[0097] See also Figure 2 As shown, an embodiment of the present application provides a private cloud resource demand prediction and configuration system based on spatiotemporal sequence. The private cloud resource demand prediction and configuration system based on spatiotemporal sequence includes:

[0098] A historical data acquisition and preprocessing module is used to acquire the cloud resource configuration historical data of all private cloud management domains based on the cloud resource configuration historical logs of the respective private cloud management domains, and preprocess the cloud resource configuration historical data;

[0099] A sample data generation module is used to perform periodic change identification processing based on the cloud resource configuration historical data that has completed the preprocessing, and extract corresponding prediction sample data;

[0100] A cloud resource configuration demand prediction module is used to train a spatiotemporal sequence model based on the prediction sample data, and to predict the cloud resource configuration demand data of all private cloud management domains within a preset future time period based on the trained spatiotemporal sequence model;

[0101] A cloud resource configuration characteristic information determination module is used to determine the cloud resource configuration characteristic information corresponding to each private cloud management domain based on the cloud resource configuration requirement data;

[0102] A management domain cluster generation module is used to divide all private cloud management domains with similar cloud resource configuration feature information from all private cloud management domains according to different regions and business types based on the cloud resource configuration feature information corresponding to each private cloud management domain, thereby forming corresponding management domain clusters;

[0103] The cloud resource configuration adjustment module is configured to compare and analyze the future cloud resource demand information of each private cloud management domain with the configuration utilization rate threshold, and optimize and adjust all cloud resources under the private cloud management domain.

[0104] The private cloud resource demand prediction and configuration method based on space-time sequence has the following advantages. The cloud resource configuration historical data of all private cloud management domains is obtained from the cloud resource configuration historical logs of the private cloud management domains, the resource configuration historical data is preprocessed and periodically changed, the corresponding prediction sample data is extracted, and sufficient data is provided for subsequent training of the space-time sequence model. Based on the trained space-time sequence model, the cloud resource configuration demand data of all private cloud management domains in a future preset time period is predicted, the cloud resource configuration feature information corresponding to each private cloud management domain is determined, and the same cloud resource configuration feature information is divided from all private cloud management domains according to different regions and business types, to form a corresponding management domain cluster. The historical data and data characteristics of the management domain cluster are used to predict the cloud resources of the management domains belonging to different management domain clusters. The future cloud resource demand information of each private cloud management domain is compared and analyzed with the configuration utilization rate threshold, and all cloud resources under the private cloud management domain are optimized and adjusted to improve the utilization efficiency of the cloud resources.

[0105] In another embodiment, the historical data acquisition and preprocessing module is configured to acquire cloud resource configuration historical data of all private cloud management domains based on cloud resource configuration historical logs of the private cloud management domains, and preprocess the cloud resource configuration historical data, including:

[0106] The cloud resource configuration historical logs of the private cloud management domains are analyzed to obtain all cloud resource configuration events of each private cloud management domain in a preset historical time interval. The cloud resource configuration process of all cloud resource configuration events is analyzed to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event and the configuration event occurrence time, which are used as the cloud resource configuration historical data. The cloud resource configuration historical data is sequentially subjected to interference noise data cleaning preprocessing and missing data filling preprocessing.

[0107] The sample data generation module is configured to perform periodic change identification processing on the preprocessed cloud resource configuration historical data to extract corresponding prediction sample data, including:

[0108] The cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain the periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the bad point data of the cloud resource configuration historical data that has completed all preprocessing is eliminated to obtain the corresponding prediction sample data.

[0109] The beneficial effects of the above embodiment are that the private cloud management domain contains different types of cloud resources, which may include but are not limited to computing power cloud resources, storage cloud resources, and network bandwidth cloud resources. When a terminal performs a complex computing task, it needs to rely on the cloud resources under the private cloud management domain to undertake the corresponding computing task. In this way, the private cloud management domain will generate a corresponding cloud resource configuration history log. The cloud resource configuration history logs of several private cloud management domains are analyzed to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval. The cloud resource configuration process of all cloud resource configuration events is analyzed separately to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time of occurrence of the configuration event, which is used as the cloud resource configuration history data. This ensures that the cloud resource configuration history data fully and accurately reflects the actual cloud resource configuration status of the private cloud management domain. The cloud resource configuration history data is then subjected to interference noise data cleaning preprocessing and missing data filling preprocessing, which can improve the accuracy and credibility of the cloud resource configuration history data. Then, the cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain the periodic change trend characteristic information of the cloud resource configuration historical data. In this way, the bad point data of the cloud resource configuration historical data that has completed all preprocessing is eliminated to obtain the corresponding prediction sample data, thereby improving the quality and reliability of the prediction sample data.

[0110] In another embodiment, the cloud resource configuration history data is sequentially preprocessed to fill in missing data, including:

[0111] Extract two valid data adjacent to and before the data position corresponding to the missing data as the first valid data and the second valid data;

[0112] Acquire a data difference between the first valid data and the second valid data according to the first valid data and the second valid data;

[0113] When the data difference between the first valid data and the second valid data is 0, the first valid data and the second valid data are weighted averaged to obtain a weighted average, and the weighted average is used as the first filling data to fill the missing data;

[0114] When the data difference between the first valid data and the second valid data is not 0, the data difference between the first valid data and the second valid data is used as a difference threshold;

[0115] Retrieving, from the historical data, historical data whose difference with the first valid data exceeds the difference threshold as the first historical data;

[0116] Retrieving, from the historical data, historical data whose difference with the second valid data exceeds the difference threshold as the second historical data;

[0117] The second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula:

[0118] Wherein, X represents the second filling data; X p represents the average value of valid data of historical data of cloud resource configuration; n represents the number of data of the first historical data; X 01i represents the data value corresponding to the first historical data of the i-th item; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02p represents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0119] Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data;

[0120] The missing data is filled using the second filling data.

[0121] The beneficial effect of the above embodiment is that by filling missing data, the integrity of historical cloud resource configuration data is ensured, which is crucial for subsequent data analysis and decision-making. Missing data may affect the accuracy of data analysis, and this technical solution can effectively address this issue. The technical solution adopts two different filling strategies: a weighted average is used when the difference between adjacent valid data is zero, and a more complex calculation based on historical data is performed when the difference is not zero. This strategy makes the filling process more flexible, allowing the most appropriate filling method to be selected according to different situations. When the difference between adjacent valid data is not zero, the technical solution calculates the difference between the first and second historical data and the valid data, as well as statistics such as the standard deviation and mean, to derive a more accurate filling value (the second filling data). This method considers the influence of multiple historical data points, thereby improving the accuracy of the filling. An adjustment coefficient is introduced to improve adaptability: The technical solution introduces an adjustment coefficient s, which is calculated based on the number of data points, the standard deviation, and the difference between the second historical data and the first and second valid data. The introduction of the adjustment coefficient makes the filling process more flexible and adaptable, and can better handle missing data filling problems in different situations. The entire technical solution forms a complete data processing flow, starting with extracting adjacent valid data, calculating data differences, acquiring historical data, and calculating fill values. This flow is not only logically clear but also easy to implement, capable of efficiently processing large amounts of historical cloud resource configuration data.

[0122] In summary, this technical solution improves the integrity and accuracy of data by accurately filling in the missing data in the historical data of cloud resource configuration, enhances the flexibility and adaptability of data processing, optimizes the data processing process, and provides strong support for subsequent data analysis and decision-making.

[0123] In another embodiment, the cloud resource configuration demand prediction module is used to train a spatiotemporal sequence model based on the prediction sample data, and predict the cloud resource configuration demand data of all private cloud management domains in a future preset time period based on the trained spatiotemporal sequence model, including:

[0124] The predicted sample data is input into a Graph WaveNet model to train the Graph WaveNet model. Based on the trained Graph WaveNet model, cloud resource configuration demand data for all private cloud management domains in a future preset time period is predicted. The cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain in each unit time interval in the future preset time period.

[0125] The cloud resource configuration characteristic information determination module is used to determine the cloud resource configuration characteristic information corresponding to each private cloud management domain based on the cloud resource configuration requirement data, including:

[0126] Analyze the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period;

[0127] The management domain cluster generation module is used to divide all private cloud management domains with similar cloud resource configuration feature information from all private cloud management domains according to different regions and business types based on the cloud resource configuration feature information corresponding to each private cloud management domain, thereby forming corresponding management domain clusters, including:

[0128] Based on the demand period distribution characteristic information corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristic information, region and business type are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters.

[0129] The beneficial effects of the above embodiment include inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model. Based on the trained Graph WaveNet model, cloud resource configuration demand data for all private cloud management domains within a future preset time period is predicted. This allows for a comprehensive and detailed prediction of each private cloud management domain's cloud resource demand within the future preset time period. This cloud resource configuration demand data is also analyzed to obtain characteristic information about the time period distribution of each private cloud management domain's demand for different types of cloud resources, thereby quantifying the cloud resource demand of each private cloud management domain. Based on this characteristic information about the time period distribution of each private cloud management domain's demand for different types of cloud resources, all private cloud management domains with complementary cloud resource configuration relationships are selected from all private cloud management domains to form a corresponding management domain cluster. This ensures that all private cloud management domains within the management domain cluster can uniformly configure at least one type of cloud resource across all unit time intervals within the future preset time period, providing stable and reliable cloud resource adjustment services for corresponding terminals.

[0130] In another embodiment, the cloud resource configuration adjustment module is configured to compare and analyze the future cloud resource demand information of each private cloud management domain with the configured utilization rate threshold, and optimize and adjust all cloud resources under the private cloud management domain, including:

[0131] The future cloud resource demand information of each private cloud management domain and the corresponding configuration utilization rate thresholds of all types of cloud resources are compared and analyzed to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time intervals.

[0132] The beneficial effect of the above embodiment is to compare and analyze the future cloud resource demand information of each private cloud management domain and the configuration utilization rate thresholds corresponding to all types of cloud resources, and obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period. In this way, all cloud resources under the private cloud management domain are expanded and adjusted in the corresponding time interval, which can take into account the continuous and stable cloud resource adjustment of different management domains and the targeted guarantee that the management domain obtains sufficient cloud resource quantity configuration.

[0133] In general, the method and system for predicting and configuring private cloud resource demand based on spatiotemporal sequences obtains the cloud resource configuration historical data of all private cloud management domains from their respective cloud resource configuration historical logs, preprocesses the resource configuration historical data and identifies periodic changes, and extracts corresponding prediction sample data to provide sufficient data for subsequent training of the spatiotemporal sequence model. Based on the trained spatiotemporal sequence model, the method and system predicts the cloud resource configuration requirements of all private cloud management domains in a preset future time period, thereby determining the cloud resource configuration feature information corresponding to each private cloud management domain. Furthermore, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters. The historical data and data features of the management domain clusters are used to perform cloud resource prediction with strong characteristics for management domains belonging to different management domain clusters. Furthermore, based on the future cloud resource demand information of each private cloud management domain, a comparative analysis is performed with the configuration utilization rate threshold, and all cloud resources under the private cloud management domain are optimized and adjusted to improve the utilization efficiency of cloud resources.

[0134] The above is only a specific embodiment of the present invention, and any other improvements made based on the concept of the present invention are considered to be within the protection scope of the present invention.

Claims

1. A private cloud resource demand prediction and configuration method based on spatiotemporal sequences, characterized in that: include: Based on the cloud resource configuration history logs of the respective private cloud management domains, obtaining the cloud resource configuration history data of all the private cloud management domains, and preprocessing the cloud resource configuration history data; Perform periodic change identification processing based on the pre-processed cloud resource configuration historical data to extract corresponding prediction sample data; Based on the predicted sample data, a spatiotemporal sequence model is trained, and based on the trained spatiotemporal sequence model, cloud resource configuration demand data for all private cloud management domains in a future preset time period is predicted; based on the cloud resource configuration demand data, cloud resource configuration feature information corresponding to each private cloud management domain is determined; based on the cloud resource configuration feature information corresponding to all private cloud management domains, all private cloud management domains with similar cloud resource configuration feature information are divided from all private cloud management domains according to different regions and business types to form corresponding management domain clusters, including: Inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model, and predicting cloud resource configuration demand data for all private cloud management domains within a future preset time period based on the trained Graph WaveNet model; wherein the cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain for each unit time interval within the future preset time period; Analyzing the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period; Based on the demand period distribution characteristics corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristics, regions, and business types are divided from all private cloud management domains to form corresponding management domain clusters according to different regions and business types. Based on the cloud resource configuration demand data of each private cloud management domain, a comparison analysis is performed with the configuration utilization threshold, and all cloud resources under the private cloud management domain are optimized and adjusted.

2. The method for predicting and configuring private cloud resource requirements based on spatiotemporal sequences according to claim 1, wherein: Based on the cloud resource configuration history logs of the respective private cloud management domains, obtaining the cloud resource configuration history data of all the private cloud management domains, and preprocessing the cloud resource configuration history data; Based on the pre-processed cloud resource configuration historical data, periodic change identification processing is performed to extract corresponding prediction sample data, including: Analyze the cloud resource configuration history logs of each of several private cloud management domains to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval; perform cloud resource configuration process analysis on all cloud resource configuration events to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time when the configuration event occurred, as the cloud resource configuration history data; and perform interference noise data cleaning preprocessing and missing data filling preprocessing on the cloud resource configuration history data. The cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the cloud resource configuration historical data that has completed all preprocessing is subjected to bad point data elimination processing to obtain corresponding prediction sample data.

3. The method for predicting and configuring private cloud resource requirements based on spatiotemporal sequences according to claim 2, wherein: The cloud resource configuration historical data is sequentially preprocessed to fill in missing data, including: Extract two valid data adjacent to and before the data position corresponding to the missing data as the first valid data and the second valid data; Acquire a data difference between the first valid data and the second valid data according to the first valid data and the second valid data; When the data difference between the first valid data and the second valid data is 0, the first valid data and the second valid data are weighted averaged to obtain a weighted average, and the weighted average is used as the first filling data to fill the missing data; When the data difference between the first valid data and the second valid data is not 0, the data difference between the first valid data and the second valid data is used as a difference threshold; Retrieving, from the historical data, historical data whose difference with the first valid data exceeds the difference threshold as the first historical data; Retrieving, from the historical data, historical data whose difference with the second valid data exceeds the difference threshold as the second historical data; The second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula: Wherein, X represents the second filling data; X p represents the average value of valid data of historical data of cloud resource configuration; n represents the number of data of the first historical data; X 01i represents the data value corresponding to the first historical data of the i-th item; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02p represents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula: Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data; The missing data is filled using the second filling data.

4. The method for predicting and configuring private cloud resource requirements based on spatiotemporal sequences according to claim 3, wherein: Based on the cloud resource configuration demand data of each private cloud management domain, the cloud resource configuration demand data is compared with the configuration utilization threshold, and all cloud resources under the private cloud management domain are optimized and adjusted, including: Compare and analyze the cloud resource configuration demand data of each private cloud management domain and the configuration utilization rate thresholds corresponding to all types of cloud resources to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, expand and adjust all cloud resources under the private cloud management domain in the corresponding time interval.

5. A private cloud resource demand prediction and configuration system based on spatiotemporal sequences, characterized by: include: A historical data acquisition and preprocessing module is used to acquire the cloud resource configuration historical data of all private cloud management domains based on the cloud resource configuration historical logs of the respective private cloud management domains, and preprocess the cloud resource configuration historical data; A sample data generation module is used to perform periodic change identification processing based on the pre-processed cloud resource configuration historical data, and extract corresponding prediction sample data; The cloud resource configuration demand prediction module is used to train the spatiotemporal sequence model based on the prediction sample data, and predict the cloud resource configuration demand data of all private cloud management domains in a preset future time period based on the trained spatiotemporal sequence model, including: Inputting the predicted sample data into a Graph WaveNet model to train the Graph WaveNet model, and predicting cloud resource configuration demand data for all private cloud management domains within a future preset time period based on the trained Graph WaveNet model; wherein the cloud resource configuration demand data includes the type and quantity of cloud resources required by each private cloud management domain for each unit time interval within the future preset time period; A cloud resource configuration characteristic information determination module is configured to determine the cloud resource configuration characteristic information corresponding to each private cloud management domain based on the cloud resource configuration requirement data, including: Analyzing the cloud resource configuration demand data to obtain time period distribution characteristic information of the demand for different types of cloud resources by each private cloud management domain; wherein the time period distribution characteristic information refers to the distribution information of all unit time intervals corresponding to the demand for each type of cloud resource by each private cloud management domain within the future preset time period; The management domain cluster generation module is used to divide all private cloud management domains with similar cloud resource configuration feature information from all private cloud management domains according to different regions and business types based on the cloud resource configuration feature information corresponding to each private cloud management domain, thereby forming corresponding management domain clusters, including: Based on the demand period distribution characteristics corresponding to all private cloud management domains, all private cloud management domains with the same demand period distribution characteristics, regions, and business types are divided from all private cloud management domains to form corresponding management domain clusters according to different regions and business types. The cloud resource configuration adjustment module is used to compare and analyze the cloud resource configuration demand data of each private cloud management domain with the configuration utilization rate threshold, and optimize and adjust all cloud resources under the private cloud management domain.

6. The private cloud resource demand prediction and configuration system based on spatiotemporal sequence according to claim 5, characterized in that: The historical data acquisition and preprocessing module is used to acquire the cloud resource configuration historical data of all private cloud management domains based on the cloud resource configuration historical logs of the respective private cloud management domains, and preprocess the cloud resource configuration historical data, including: Analyze the cloud resource configuration history logs of each of several private cloud management domains to obtain all cloud resource configuration events that occurred in each private cloud management domain within a preset historical time interval; perform cloud resource configuration process analysis on all cloud resource configuration events to obtain the cloud resource type and quantity corresponding to each cloud resource configuration event, as well as the time when the configuration event occurred, as the cloud resource configuration history data; and perform interference noise data cleaning preprocessing and missing data filling preprocessing on the cloud resource configuration history data. The sample data generation module is used to perform periodic change identification processing based on the pre-processed cloud resource configuration historical data to extract corresponding prediction sample data, including: The cloud resource configuration historical data that has completed all preprocessing is subjected to three exponential smoothing processes to obtain periodic change trend characteristic information of the cloud resource configuration historical data; based on the periodic change trend characteristic information, the cloud resource configuration historical data that has completed all preprocessing is subjected to bad point data elimination processing to obtain corresponding prediction sample data.

7. The private cloud resource demand prediction and configuration system based on spatiotemporal sequence according to claim 6, characterized in that: The cloud resource configuration historical data is sequentially preprocessed to fill in missing data, including: Extract two valid data adjacent to and before the data position corresponding to the missing data as the first valid data and the second valid data; Acquire a data difference between the first valid data and the second valid data according to the first valid data and the second valid data; When the data difference between the first valid data and the second valid data is 0, the first valid data and the second valid data are weighted averaged to obtain a weighted average, and the weighted average is used as the first filling data to fill the missing data; When the data difference between the first valid data and the second valid data is not 0, the data difference between the first valid data and the second valid data is used as a difference threshold; Retrieving, from the historical data, historical data whose difference with the first valid data exceeds the difference threshold as the first historical data; Retrieving, from the historical data, historical data whose difference with the second valid data exceeds the difference threshold as the second historical data; The second filling data is obtained by using the first historical data and the second historical data; wherein the second filling data is obtained by the following formula: Wherein, X represents the second filling data; X p represents the average value of valid data of historical data of cloud resource configuration; n represents the number of data of the first historical data; X 01i represents the data value corresponding to the first historical data of the i-th item; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02p represents the average value of the data values ​​corresponding to the m second historical data; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula: Where s represents the adjustment coefficient; m represents the number of data in the second historical data; X 01 Indicates the data value corresponding to the first valid data; X 02 Indicates the data value corresponding to the second valid data; X 01b represents the standard deviation of the data values ​​corresponding to the n first historical data; X 02b represents the standard deviation of the data values ​​corresponding to the m second historical data; X 02i represents the data value corresponding to the i-th second historical data; The missing data is filled using the second filling data.

8. The private cloud resource demand prediction and configuration system based on spatiotemporal sequence according to claim 7, characterized in that: The cloud resource configuration adjustment module is used to compare and analyze the cloud resource configuration demand data of each private cloud management domain with the configuration utilization threshold, and optimize and adjust all cloud resources under the private cloud management domain, including: Compare and analyze the cloud resource configuration demand data of each private cloud management domain and the configuration utilization rate thresholds corresponding to all types of cloud resources to obtain the allowed configuration time intervals for all types of cloud resources required by each private cloud management domain in the future time period; then, based on the allowed configuration time intervals, expand and adjust all cloud resources under the private cloud management domain in the corresponding time interval.

Images