Method for training data resource prediction model of cloud server and resource prediction method

Abstract

The application discloses a cloud server data resource prediction model training method and resource prediction method, a storage medium, equipment and a computer program product, and belongs to the data resource prediction field. The method comprises the following steps: acquiring a resource sequence of a cloud server; decomposing the resource sequence into a plurality of wavelet sequences to obtain sequence characteristics of each resource sequence; and training a bidirectional long short-term memory recurrent neural network according to the sequence characteristics of the resource sequence to obtain a data resource prediction model. The wavelet decomposition method is adopted to decompose the resource sequence into a plurality of characteristic components with different frequency intervals, so that multi-level information of resource load change is captured, the representativeness of training data and the generalization ability of the model are improved, the time series modeling capability of the multi-layer bidirectional long short-term memory network is improved, and the problem of insufficient data resource prediction accuracy of the cloud server in the related art is solved.

Description

Technical Field

[0001] This application belongs to the field of data resource prediction, specifically relating to a training method for a data resource prediction model of a cloud server, a resource prediction method, a storage medium, a device, and a computer program product. Background Technology

[0002] With the rapid development of cloud computing technology, more and more enterprises and organizations are migrating their computing, storage, and network resources to cloud platforms. As a core component of cloud computing, cloud servers handle a large number of user requests and data processing tasks, making the rational allocation and management of their resources particularly important. To ensure that cloud servers can efficiently utilize resources while providing stable and high-performance services, accurately predicting future resource needs has become a key technology.

[0003] Among existing cloud server resource prediction methods, a common approach is to combine a multi-layer Bidirectional Long Short-Term Memory (BiLSTM) network with an improved Particle Swarm Optimization (PSO) algorithm. This method predicts application load by constructing a multi-layer BiLSTM neural network and uses the PSO algorithm to optimize the hyperparameters of the neural network to improve the model's prediction accuracy. Furthermore, a multi-index fusion method based on benchmark models is used to evaluate the prediction model to ensure its stability and reliability.

[0004] However, the above methods suffer from insufficient accuracy and stability in predicting results. Summary of the Invention

[0005] This application aims to provide a training method for a data resource prediction model of a cloud server, as well as a resource prediction method, storage medium, device, and computer program product, which at least solves the problem of insufficient accuracy in data resource prediction for cloud servers.

[0006] In a first aspect, embodiments of this application disclose a method for training a data resource prediction model for a cloud server, comprising:

[0007] Obtain the resource sequence of the cloud server; the resource sequence is used to characterize the change of the resource load of the cloud server over time;

[0008] The resource sequence is decomposed into multiple wavelet sequences to obtain the sequence features of each resource sequence; the sequence features contain multiple feature components, each feature component corresponding to a different frequency range; the feature components are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature component.

[0009] The bidirectional long short-term memory recurrent neural network is trained based on the sequence characteristics of the resource sequence to obtain a data resource prediction model; the resource prediction model is used to predict the resource load of the cloud server based on the resource sequence to obtain the sequence prediction characteristics of the cloud server.

[0010] Secondly, embodiments of this application also disclose a method for predicting data resources of a cloud server, including:

[0011] According to the preset number of predicted samplings, obtain the resource sequences of multiple cloud servers;

[0012] The resource sequences are respectively input into the data resource prediction model to obtain multiple sequence prediction features of the cloud server; the data resource prediction model is trained according to the method described in the first aspect.

[0013] Based on the obtained sequence prediction features, the predicted value of the resource requirement for the cloud server is determined.

[0014] Thirdly, embodiments of this application also disclose a training apparatus for a data resource prediction model of a cloud server, comprising:

[0015] The sequence module is used to obtain the resource sequence of the cloud server; the resource sequence is used to characterize the change of the resource load of the cloud server over time.

[0016] The decomposition module is used to decompose the resource sequence into multiple wavelet sequences to obtain the sequence features of each resource sequence; the sequence features include multiple feature components, each feature component corresponding to a different frequency range; the feature components are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature component.

[0017] The training module is used to train the bidirectional long short-term memory recurrent neural network based on the sequence characteristics of the resource sequence to obtain a data resource prediction model; the resource prediction model is used to predict the resource load of the cloud server based on the resource sequence to obtain the sequence prediction characteristics of the cloud server.

[0018] Fourthly, embodiments of this application also disclose a data resource prediction device for a cloud server, comprising:

[0019] The sampling module is used to obtain resource sequences of multiple cloud servers according to a preset number of predicted sampling times;

[0020] The prediction module is used to input the resource sequences into the data resource prediction model to obtain multiple sequence prediction features of the cloud server; the data resource prediction model is trained according to the method described in the first aspect.

[0021] An estimation module is used to determine the predicted value of the resource requirements of the cloud server based on the obtained multiple sequence prediction features.

[0022] Fifthly, embodiments of this application also disclose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the first or second aspect.

[0023] Sixthly, embodiments of this application also disclose an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, performs the steps as described in the first or second aspect.

[0024] In a seventh aspect, embodiments of this application also disclose a computer program product, on which a computer program is stored, and when the computer program is executed by a processor, it performs the steps as described in the first or second aspect.

[0025] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies. Attached Figure Description

[0026] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0027] Figure 1This is a flowchart illustrating the steps of a training method for a data resource prediction model for a cloud server, as provided in an embodiment of this application.

[0028] Figure 2 This is a schematic diagram representing the resource sequence in the embodiments of this application;

[0029] Figure 3 A schematic diagram of a wavelet function decomposition process in an embodiment of this application;

[0030] Figure 4 This is a flowchart of the steps of training a data resource prediction model for a cloud server provided in an embodiment of this application;

[0031] Figure 5 This is a flowchart illustrating the steps of a cloud server data resource prediction method provided in an embodiment of this application;

[0032] Figure 6 This is a flowchart of another cloud server data resource prediction method provided in this application embodiment;

[0033] Figure 7 This is a comparison chart of experimental results obtained according to the method of the embodiments of this application;

[0034] Figure 8 This is a flowchart of a resource estimation process obtained according to the method of the embodiments of this application;

[0035] Figure 9 This is a flowchart of a program execution process obtained according to the method of the embodiments of this application;

[0036] Figure 10 This is a schematic diagram of the structure of a training device for a data resource prediction model of a cloud server provided in an embodiment of this application;

[0037] Figure 11 This is a schematic diagram of the structure of a data resource prediction device for a cloud server provided in an embodiment of this application;

[0038] Figure 12 This is a block diagram of an electronic device provided in an embodiment of this application;

[0039] Figure 13 This is a block diagram of another electronic device provided in the embodiments of this application. Detailed Implementation

[0040] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0041] To address the problems existing in current technologies, this solution considers that low-frequency components typically contain the overall trend and long-term changes in the data, while high-frequency components mainly contain the details and short-term fluctuations. This separation enables the model to more clearly understand and process the different features in the data, improving the accuracy and stability of predictions.

[0042] Furthermore, because wavelet transform can perform analysis simultaneously in both the time and frequency domains, it is used to extract both types of components. This method decomposes the signal into components of different scales, generating detail signals (high-frequency components) and approximate signals (low-frequency components), thus comprehensively capturing both the overall trend and local details of the signal. The multi-resolution analysis and the ability to simultaneously provide time and frequency information enable wavelet transform to accurately extract and process multi-level features in signals, improving the accuracy and stability of prediction models.

[0043] Based on this Figure 1 This embodiment provides a training method for a data resource prediction model for a cloud server, which may include the following steps:

[0044] Step 101: Obtain the resource sequence of the cloud server.

[0045] Among them, the resource sequence is used to characterize the change of cloud server resource load over time.

[0046] In some embodiments of this application, obtaining the resource sequence of the cloud server is to comprehensively capture the changes in the cloud server's resource load over time. The specific process includes extracting resource data from the cloud server to characterize the changes in the cloud server's resource load over time. The resource sequence is used to characterize the changes in the cloud server's resource load over time. This process can provide basic data for subsequent resource prediction models, thereby effectively improving the accuracy of predictions.

[0047] like Figure 2As shown, in a specific example, researchers collected Central Processing Unit (CPU) utilization data from cloud servers deployed on the OpenStack platform and recorded resource sequence data at regular intervals. By continuously collecting this resource data, researchers were able to accurately depict the changes in cloud server resource load and provide reliable foundational data for subsequent wavelet decomposition and bidirectional long short-term memory recurrent neural network training. Following this process, researchers ultimately obtained a complete set of resource sequences, providing necessary support for subsequent model training and resource prediction.

[0048] Step 102: Decompose the resource sequence into multiple wavelet sequences to obtain the sequence features of each resource sequence.

[0049] The sequence features contain multiple feature components, each corresponding to a different frequency range. The feature components are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature component.

[0050] In some embodiments of this application, the resource sequence is decomposed into multiple wavelet sequences to extract multi-level variation features of cloud server resource load. This process involves performing a wavelet transform on the original resource sequence to generate multiple wavelet sequences. Wavelet transform (WT) is a signal processing method that can simultaneously provide time and frequency information, decomposing a signal into a combination of approximate and detailed signals. Through wavelet decomposition, the variation trend and detailed features of resource load can be effectively separated, obtaining multiple feature components of the sequence features, each corresponding to a different frequency range. This step allows the model to better understand and capture changes in resource load, thereby providing more comprehensive and detailed data support for subsequent model training.

[0051] like Figure 3 As shown, in a specific example, researchers used the Daubechies wavelet function to perform a three-level wavelet decomposition on the CPU utilization sequence of a cloud server. This process generates three detail components and one approximation component. The approximation component preserves low-frequency information and overall trends of resource load, while the detail component captures high-frequency details. Using this method, researchers can effectively extract multi-level features from the resource sequence and provide rich feature data for subsequent BiLSTM model training. Ultimately, researchers obtained a detailed set of sequence features that provide crucial support for model training and resource prediction.

[0052] Step 103: Train the bidirectional long short-term memory recurrent neural network based on the sequence characteristics of the resource sequence to obtain a data resource prediction model.

[0053] Among them, the resource prediction model is used to predict the resource load of cloud servers based on resource sequences in order to obtain the sequence prediction characteristics of cloud servers.

[0054] In some embodiments of this application, training a BiLSTM based on the sequence characteristics of the resource sequence aims to capture the temporal features of cloud server resource load, thereby improving the accuracy and stability of predictions. Specifically, the process involves inputting the extracted sequence features into the BiLSTM model for training to obtain a data resource prediction model. BiLSTM is a neural network that combines LSTM and bidirectional properties, enabling it to capture long-term dependencies in data from both forward and backward dimensions. This process, by fully mining the temporal dependencies in the data, enables the model to possess higher predictive capabilities, thereby obtaining the sequence prediction features of the cloud server.

[0055] In a specific example, researchers input multi-level feature components of cloud server CPU utilization extracted through wavelet transform into a BiLSTM model for training. During training, the model continuously adjusts its parameters to minimize prediction error and gradually optimizes its predictive ability. Through this training method, researchers ultimately obtained a BiLSTM data resource prediction model capable of accurately predicting cloud server resource load. This model can not only capture the overall trend of resource load changes but also identify short-term fluctuations, thus providing accurate prediction results in practical applications.

[0056] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies.

[0057] Figure 4This embodiment provides another method for training a cloud server data resource prediction model, which may include the following steps:

[0058] Step 201: Obtain the resource sequence of the cloud server.

[0059] Among them, the resource sequence is used to characterize the change of cloud server resource load over time.

[0060] The method shown in this step has been explained in step 101 and will not be repeated here.

[0061] Step 202: Decompose the resource sequence into multiple wavelet sequences to obtain the sequence features of each resource sequence.

[0062] The sequence features contain multiple feature components, each corresponding to a different frequency range. The feature components are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature component.

[0063] The method shown in this step has been explained in step 102 and will not be repeated here.

[0064] Optionally, step 202 includes the following sub-steps:

[0065] Sub-step 2021: Decompose the resource sequence into a first resource subsequence and a second resource subsequence.

[0066] Among them, the frequency value in the frequency range corresponding to the first resource subsequence is higher than the frequency value in the frequency range corresponding to the second resource subsequence.

[0067] In some embodiments of this application, decomposing the resource sequence into a first resource subsequence and a second resource subsequence is to separate the high-frequency and low-frequency information of the resource load. Specifically, this process involves using wavelet transform to decompose the resource sequence according to frequency range, obtaining the first and second resource subsequences. The first resource subsequence contains higher-frequency components, while the second resource subsequence contains lower-frequency components. This decomposition allows for better extraction of different frequency features from the resource sequence, providing more accurate data for subsequent model training.

[0068] In a specific example, researchers performed wavelet transform decomposition on CPU utilization time-series data obtained from a cloud server. They chose the Daubechies wavelet function to decompose the data, generating a detail component (the first resource subsequence) and an approximate component (the second resource subsequence). The detail component contains high-frequency information, while the approximate component contains low-frequency information. Through this decomposition method, the researchers successfully decomposed the original resource sequence into subsequences containing different frequency features, providing a foundation for subsequent analysis and processing. Ultimately, they obtained a detailed set of first and second resource subsequences, laying the groundwork for training a subsequent resource prediction model.

[0069] In sub-step 2022, if the number of times the resource sequence is decomposed into the first resource subsequence and the second resource subsequence is less than the preset decomposition iteration threshold, the first resource subsequence obtained by decomposition is recorded as a detail sequence, the resource sequence is updated with the second resource subsequence, and the process returns to sub-step 2021.

[0070] In some embodiments of this application, if the number of times the resource sequence is decomposed into a first resource subsequence and a second resource subsequence is less than a preset decomposition iteration threshold, the first resource subsequence obtained from the decomposition is recorded as a detail sequence, and the resource sequence is updated with the second resource subsequence. This is to gradually decompose the high-frequency and low-frequency information in the resource load. Specifically, after each decomposition, the first resource subsequence is recorded as a detail sequence, and the current resource sequence is updated with the second resource subsequence. Then, the decomposition is performed again until the preset decomposition iteration threshold is reached. The detail sequence contains higher-frequency components, while the updated resource sequence contains lower-frequency components. Through this gradual decomposition method, multi-level features of the resource load can be fully extracted, improving the model's predictive ability.

[0071] In a specific example, researchers performed wavelet decomposition on time-series data of cloud server CPU utilization. Assuming a preset decomposition iteration threshold of three iterations, the researchers obtained a first resource subsequence (containing high-frequency information) and a second resource subsequence (containing low-frequency information) during the first decomposition. The first resource subsequence was recorded as a detail sequence, and the current resource sequence was updated using the second resource subsequence. Then, the researchers continued to perform a second and third decomposition on the updated resource sequence, repeating the process of recording the detail sequence and updating the resource sequence. Ultimately, through three iterations, the researchers obtained the complete detail sequence and approximate sequence, providing sufficient data support for subsequent analysis and processing.

[0072] Sub-step 2023: When the number of executions reaches the decomposition iteration threshold, the second resource sub-sequence obtained by decomposition is recorded as an approximate sequence, and the sequence characteristics of the resource sequence data are determined based on all the detailed sequences and the approximate sequence.

[0073] In some embodiments of this application, when the number of decomposition iterations reaches a threshold, the second resource subsequence obtained from the decomposition is recorded as an approximate sequence in order to obtain the low-frequency features of the resource sequence. Specifically, after each decomposition, the obtained second resource subsequence is recorded as an approximate sequence. The detail sequence represents high-frequency information, while the approximate sequence represents low-frequency information. Using this method, the complete sequence features of the resource sequence data can ultimately be determined based on all the detail sequences and approximate sequences, thereby providing rich temporal information for the model and improving prediction accuracy.

[0074] In a specific example, researchers performed wavelet decomposition on time-series data of cloud server CPU utilization, setting the decomposition iteration threshold to three times. After three decompositions, the researchers recorded the detailed sequences obtained from each decomposition and recorded the second resource subsequence obtained from the last decomposition as an approximate sequence. Finally, based on all recorded detailed and approximate sequences, the researchers comprehensively determined the multi-level sequence features of the resource sequence data. Using this method, the researchers were able to effectively extract the full picture of cloud server resource load, providing sufficient data support for subsequent model training and prediction. Ultimately, the researchers obtained detailed sequence features, laying the foundation for high-precision resource demand prediction.

[0075] Optionally, in order to determine the sequence characteristics of the resource sequence data based on all detailed sequences and approximate sequences, sub-step 2023 includes the following sub-steps:

[0076] Sub-step 20231: Determine all detailed sequences and approximate sequences as resource subsequences of the resource sequence, and normalize each resource subsequence to obtain the resource normalized subsequence corresponding to the cloud server and each frequency range.

[0077] In some embodiments of this application, all detail sequences and approximate sequences are each determined as a resource subsequence of the resource sequence, and each resource subsequence is normalized to eliminate the magnitude differences between different feature data and ensure the stability of the model. Specifically, the process involves treating the decomposed detail sequences and approximate sequences as independent resource subsequences and normalizing each resource subsequence. Normalization maps the data to a standard range, such as the [0,1] interval. This process reduces the weight differences between data features, improving the stability and robustness of the prediction model.

[0078] In a specific example, researchers used wavelet decomposition to obtain detailed and approximate sequences of cloud server CPU utilization as independent resource subsequences. They calculated the maximum and minimum values ​​for each subsequence and used these values ​​to normalize the data. Specifically, they subtracted the minimum value from the value of each subsequence and then divided it by the difference between the maximum and minimum values, ensuring that the processed data all fell within the [0,1] interval. This normalization method ensured that the characteristics of each resource subsequence were compared and analyzed on a uniform scale. Ultimately, the researchers obtained normalized resource subsequences, providing a standardized data foundation for subsequent sequence feature determination.

[0079] Sub-step 20232: According to the preset interval order, the resource normalized subsequence is used as the feature component of the sequence feature to obtain the sequence feature.

[0080] In some embodiments of this application, resource normalized subsequences are used as feature components of the sequence feature according to a preset interval order. This is to enable the model to fully utilize information from different frequency intervals for prediction. Specifically, the process involves arranging and integrating the normalized resource subsequences as different feature components of the sequence feature according to a preset interval order. By orderly integrating these feature components, a comprehensive sequence feature can be formed to reflect the changes in resource load across different frequency intervals. This method ensures that the prediction model fully captures multi-level information about resource load, thereby improving the accuracy and stability of the prediction.

[0081] In a specific example, researchers performed wavelet decomposition and normalization on the CPU utilization data of cloud servers, obtaining multiple resource-normalized subsequences. These normalized subsequences were then arranged as feature components according to a pre-defined interval order. For instance, assuming the pre-defined interval order is from low frequency to high frequency, the approximate sequence was placed first, followed by the first-level detail sequence, and so on up to the third-level detail sequence. In this way, researchers successfully obtained a complete sequence feature encompassing information from different frequency intervals, providing a sufficient data foundation for subsequent model training and prediction. Ultimately, based on these sequence features, researchers were able to more accurately predict the resource requirements of cloud servers, improving system performance and resource management efficiency.

[0082] Step 203: Sample the sequence features to obtain multiple sampled features of the cloud server.

[0083] In some embodiments of this application, sampling of sequence features is used to further extract effective features from the cloud server resource load. This process involves randomly selecting several samples from the decomposed sequence features to form multiple representative sampled features. This method allows for the extraction of more diverse and meaningful features from the original data, improving the generalization ability and prediction accuracy of subsequent models. The sampling process ensures that each feature component is fully represented in the dataset, thereby effectively reducing the model's variance and improving its stability and accuracy.

[0084] In a specific example, researchers sampled features from the cloud server CPU utilization sequence obtained after wavelet decomposition. They set a time window and randomly sampled features within each window, generating multiple training sample sets. These sample sets retained important information from the original data while increasing the diversity of the dataset through sampling. This sampling method allowed researchers to obtain a set of highly representative and diverse feature data, providing a solid foundation for subsequent model training.

[0085] Optionally, step 203 includes the following sub-steps:

[0086] Sub-step 2031: According to the preset sampling window threshold, each feature component in the sequence feature is sliced ​​into data to obtain multiple candidate features of the cloud server.

[0087] Among them, the candidate feature components of the candidate features are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature components of the candidate feature components in a preset time window.

[0088] In some embodiments of this application, data slicing is performed on each feature component in the sequence features according to a preset sampling window threshold in order to extract more detailed features from the resource load data. Specifically, the process includes slicing each feature component in the sequence features according to the preset sampling window threshold to generate multiple candidate features. Each candidate feature's feature component represents the change in the cloud server's resource load within a preset time window, corresponding to a specific frequency range. This data slicing method allows for more detailed capture of dynamic changes in resource load, improving the prediction accuracy and reliability of subsequent models.

[0089] In a specific example, researchers set a 5-second sampling window threshold to slice the sequence features of cloud server CPU utilization. Each feature component was sliced ​​according to the preset 5-second time window, generating multiple candidate features. For example, a feature component was cut into multiple small segments within the 5-second time window, each segment detailing the changes in resource load during that time period. In this way, researchers obtained a more detailed set of candidate features, providing more accurate data support for subsequent model training. Ultimately, researchers successfully generated multiple candidate features for cloud servers, laying the foundation for improving the accuracy of the prediction model.

[0090] Sub-step 2032: Use all the obtained candidate features as the sample space for the selected samples, randomly select from the sample space, and use the multiple selected candidate features as multiple sampling features of the cloud server.

[0091] In some embodiments of this application, all the obtained candidate features are used as the sample space for sampling. Random sampling from this sample space is used to generate representative sampling features from a large amount of data. Specifically, the process involves using all candidate features as the sample space and then randomly sampling several features from it as the final sampling features. This random sampling method ensures that each candidate feature has a chance to be selected, thereby increasing the diversity and representativeness of the dataset and effectively improving the model's generalization ability and prediction accuracy.

[0092] In a specific example, researchers used all candidate features obtained through data slicing as the sample space for sampling. Assuming the sample space contains 1000 candidate features, researchers could randomly select 100 features with replacement as the sampling features for the cloud server. This random sampling method ensures that the final selected sampling features are broadly representative, covering different aspects and variations of resource load data. Ultimately, researchers obtained a high-quality set of sampling features, providing a solid data foundation for subsequent model training and resource prediction. This approach allows researchers to generate diverse sampling features, improving the accuracy and robustness of the prediction model.

[0093] Step 204: Train the bidirectional long short-term memory recurrent neural network by obtaining multiple sampling features to obtain a trained data resource prediction model.

[0094] In some embodiments of this application, training the BiLSTM with multiple sampled features aims to further improve the model's predictive accuracy and stability. Specifically, this process involves inputting multiple previously sampled feature samples into the BiLSTM model for training. The BiLSTM model combines the traditional Long Short-Term Memory (LSTM) structure with bidirectional characteristics, enabling it to capture long-term dependencies in time-series data from both forward and backward dimensions. This training process allows the model to exhibit higher predictive capabilities when processing time-series data, resulting in a more accurate data resource prediction model.

[0095] In a specific example, researchers trained a BiLSTM model using multiple sampled features extracted from CPU utilization sequences of cloud servers. During training, they input several representative sampled features into the BiLSTM model, and through multiple iterations and parameter optimization, the model gradually learned the resource load variation patterns. In this way, they ultimately obtained a well-trained data resource prediction model that can accurately predict the load of cloud server resources, providing strong support for resource management and scheduling in practical applications.

[0096] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies.

[0097] In some embodiments of this application,

[0098] Figure 5 This embodiment provides a data resource prediction method for cloud servers, and the training method includes:

[0099] Step 301: Obtain the resource sequence of multiple cloud servers according to the preset number of predicted sampling times.

[0100] In some embodiments of this application, obtaining resource sequences from multiple cloud servers according to a preset number of prediction samplings is to ensure that the prediction model can learn from a sufficient number of data samples, thereby improving the accuracy and stability of the prediction. Specifically, the process includes collecting time-varying sequence data of the resource load of multiple cloud servers according to the set number of prediction samplings. This process ensures that the model has rich and diverse data samples during training and prediction, thus better capturing the changing patterns of cloud server resource load.

[0101] In a specific example, researchers collected time-series data on CPU utilization from multiple cloud servers deployed on the OpenStack platform, following a predetermined number of sampling attempts. They set a five-minute interval between each sampling and continuously collected resource data from multiple cloud servers. This method allowed them to obtain multiple sets of time-series data reflecting changes in cloud server resource load, providing ample data support for subsequent training of the data resource prediction model. Ultimately, the researchers collected a rich series of resource sequence data, laying the foundation for high-precision predictions by the model.

[0102] Step 302: Input the resource sequences into the data resource prediction model to obtain multiple sequence prediction features of the cloud server.

[0103] The data resource prediction model is trained according to the training method of the cloud server data resource prediction model in the above-disclosed embodiments.

[0104] In some embodiments of this application, resource sequences are input into a data resource prediction model to predict changes in the resource load of cloud servers based on a previously trained model. Specifically, the process involves inputting resource sequences from multiple cloud servers one by one into the pre-trained data resource prediction model to obtain multiple sequence prediction features for each cloud server. The data resource prediction model, trained in the previous steps, is used to predict the input resource sequences, thereby obtaining sequence prediction features for each cloud server. Through this process, future resource demands of cloud servers can be effectively predicted, ensuring efficient resource allocation and utilization.

[0105] In a specific example, researchers input CPU utilization sequence data from cloud servers obtained from the OpenStack platform into a previously trained BiLSTM model. Each resource sequence is processed by the model to generate corresponding sequence prediction features, which reflect changes in CPU utilization of the cloud server over a future period. In this way, researchers can obtain a detailed set of sequence prediction features for further analysis and optimization of cloud server resource management. Ultimately, based on these sequence prediction features, researchers can formulate more reasonable resource allocation strategies, improving the overall performance and stability of the system.

[0106] Step 303: Determine the predicted value of the resource requirements for the cloud server based on the obtained multiple sequence prediction features.

[0107] In some embodiments of this application, determining the predicted resource requirements of a cloud server based on multiple obtained sequence prediction features is to further quantify the resource usage of the cloud server over a future period. The specific process includes utilizing the multiple sequence prediction features obtained in previous steps, and through weighted processing and integrated analysis, ultimately determining the resource requirements of the cloud server for the future time period. In this process, the weighted processing can allocate weights based on the accuracy of each prediction model, thereby improving the accuracy of the final prediction result. Through this step, the resource requirements of the cloud server can be effectively predicted, thereby achieving optimal resource allocation and preventing resource shortages in advance.

[0108] In a specific example, researchers first obtained sequential prediction features of CPU utilization for multiple cloud servers using a BiLSTM model. Next, they calculated weighting coefficients based on the reciprocal of the error for each prediction feature and then weighted and combined the multiple predictions to arrive at the final predicted resource demand. For instance, assuming the sequential prediction features for a cloud server are Y1, Y2, and Y3, and the corresponding weighting coefficients are W1, W2, and W3, the final predicted resource demand Y can be calculated using the formula Y = W1*Y1 + W2*Y2 + W3*Y3. This method allows researchers to accurately determine the resource demand of cloud servers over a future period, providing a reliable basis for resource management and scheduling. Ultimately, the researchers obtained a series of accurate predicted resource demand values, which can effectively guide the dynamic adjustment and optimization of cloud server resources.

[0109] Furthermore, in some embodiments of this application, the weighting formula may be:

[0110] Y = W1Y1 + W2Y2 + ... + W n Y n ;

[0111] Among them, Wn Through the formula:

[0112] get,

[0113] In the formula for Wn, the i-th intermediate parameter p i Through the formula:

[0114] get;

[0115] In the above formula, X represents the actual data of the cloud server resource subsequence, and Y represents the predicted value of the cloud server resource subsequence. n W represents the predicted value of the nth BiLSTM model. n p represents the weighting coefficients of the nth BiLSTM model. n This represents the reciprocal of the error of the nth model.

[0116] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies.

[0117] Figure 6 This embodiment provides another method for predicting data resources for cloud servers.

[0118] Step 401: Obtain the resource sequence of multiple cloud servers according to the preset number of predicted sampling times.

[0119] The method shown in this step has been explained in step 301 and will not be repeated here.

[0120] Step 402: Input the resource sequences into the data resource prediction model to obtain multiple sequence prediction features of the cloud server.

[0121] The data resource prediction model is trained according to the training method of the cloud server data resource prediction model in the above-disclosed embodiments.

[0122] The method shown in this step has been explained in step 302 and will not be repeated here.

[0123] Step 403: Use the obtained multiple sequence prediction features as estimation samples to estimate the resource requirements of the cloud server, so as to determine the resource estimation sequence of the cloud server.

[0124] In some embodiments of this application, multiple obtained sequence prediction features are used as estimation samples to estimate the resource requirements of the cloud server in order to further accurately calculate the future resource usage of the cloud server. Specifically, the process involves taking multiple previously obtained sequence prediction features as input and comprehensively evaluating the resource requirements of the cloud server through an estimation model to determine the resource estimation sequence. The estimation samples represent multi-dimensional information about the cloud server's resource load. By analyzing these samples, future resource requirements can be better predicted, ensuring the effective allocation and use of resources.

[0125] In a specific example, researchers used multiple CPU utilization sequence predictions obtained through a BiLSTM model as estimation samples. These samples were then used to estimate the future resource requirements of each cloud server through a comprehensive evaluation model. Specifically, assuming the sample prediction features are S1, S2, and S3, these samples were input into the estimation model. The model analyzed each sample and generated resource requirement estimates. Finally, these estimates were integrated to obtain a complete resource estimation sequence that reflects the resource requirements of the cloud server over a future time period. This resource estimation sequence provides important information for cloud server resource management and optimization, helping to ensure stable and efficient system performance.

[0126] Optionally, step 403 includes the following sub-steps:

[0127] Sub-step 4031 involves weighted summation of the obtained multiple sequence prediction features to obtain the normalized features for cloud server resource estimation.

[0128] In some embodiments of this application, multiple sequence prediction features are weighted and summed to synthesize the prediction results of different models and obtain more accurate resource estimation normalized features. The specific process involves weighting each sequence prediction feature, with weights typically allocated based on the model's accuracy, and then summing these weighted prediction features to obtain the final resource estimation normalized features. This method fully utilizes the advantages of each prediction model, and through weighted summation, it improves the accuracy and stability of resource estimation.

[0129] In a specific example, researchers used multiple BiLSTM models to predict the CPU utilization sequence of cloud servers, with each model generating a predicted feature. The researchers calculated weights based on the reciprocal of the prediction error of each model and then performed a weighted sum of these predicted features. For example, assuming there are three predicted features Y1, Y2, and Y3, with corresponding weights W1, W2, and W3 respectively, the final resource estimation normalized feature Y can be calculated using the formula Y = W1*Y1 + W2*Y2 + W3*Y3. Through this weighted summation method, researchers can synthesize the prediction results of different models, improving the accuracy and reliability of the final prediction. Ultimately, the researchers obtained accurate resource estimation normalized features, providing a reliable data foundation for subsequent inverse normalization processing and resource estimation.

[0130] Sub-step 4032: Based on the preset normalization features, perform inverse normalization processing on the resource estimation normalization features to obtain the resource estimation features of the cloud server.

[0131] Each component of the resource estimation feature corresponds one-to-one with a frequency interval.

[0132] In some embodiments of this application, inverse normalization of the resource estimation normalization features, based on preset normalization characteristics, is performed to restore the standardized data to its original magnitude for practical application. Specifically, the process involves inverse normalizing the resource estimation normalization features based on the maximum and minimum values ​​recorded during the previous normalization process. Each component of the resource estimation feature corresponds one-to-one with a different frequency range. Through inverse normalization, the prediction results can be transformed into actual resource estimates, making the model's output more operable and practical.

[0133] In a specific example, researchers used a normalized sequence of cloud server CPU utilization for resource estimation. To perform inverse normalization, they first obtained the maximum and minimum values ​​from the normalization process. Then, they used these values ​​to perform inverse normalization on the resource estimation normalization feature. Specifically, suppose a certain resource estimation normalization feature value is Y. normalized The corresponding maximum and minimum values ​​are X, respectively. max and X min Then the resource estimation characteristic value Y after inverse normalization can be expressed by the formula Y = Y normalized *(X max -X min )+X minThe calculations yielded accurate resource estimation characteristics. Through this method, researchers were able to obtain resource estimation features that reflect actual CPU utilization, and use these features for further analysis and application. Ultimately, the researchers successfully obtained accurate resource estimation features, providing reliable data support for cloud server resource management.

[0134] Optionally, sub-step 4032 includes the following sub-steps:

[0135] Sub-step 40321 involves using the training method of the cloud server's data resource prediction model to train the data resource prediction model, and in the process of obtaining multiple sequence prediction features of the cloud server, determining the range of multiple sequence prediction features as the target range.

[0136] In some embodiments of this application, the range values ​​of multiple sequence prediction features are determined using a data resource prediction model obtained through training a data resource prediction model on a cloud server in order to quantify the fluctuation range of these prediction features. Specifically, the process includes using the trained data resource prediction model to generate multiple sequence prediction features for the cloud server and calculating the range values ​​of these features. The range value represents the difference between the maximum and minimum values ​​of the feature data. By determining these range values, clear target parameters can be provided for subsequent gain processing. Determining the range values ​​helps to understand the dispersion of the feature data, thereby providing a basis for model adjustment.

[0137] In a specific example, researchers used a previously trained BiLSTM model to generate sequence prediction features of multiple cloud server CPU utilization values. They calculated the range of each prediction feature, which is the difference between the maximum and minimum values ​​of these feature data. For example, if the maximum value of a prediction feature is 95 and the minimum value is 45, then its range is 50. In this way, the researchers successfully determined the range values ​​of each prediction feature, providing important parameters for subsequent gain processing and prediction model optimization. Ultimately, the researchers obtained a detailed set of target range values, laying the foundation for accurate resource estimation.

[0138] Sub-step 40322: Gain the normalized features according to the target range, and then inversely shift the normalized features after gaining them according to the smallest sequence prediction feature among multiple sequence prediction features to obtain resource estimation features.

[0139] In some embodiments of this application, gaining the normalized features according to the target range is intended to amplify the range of variation of the predicted features, enabling the model to better reflect resource load fluctuations. Specifically, the process involves gaining the normalized features using the previously determined target range. This gain amplifies the normalized features to a range of actual magnitude, ensuring that the predicted features reflect real resource load changes. Next, the amplified normalized features are inversely shifted according to the smallest predicted feature among multiple predicted features to obtain resource estimation features. This inverse shifting adjusts the amplified feature values ​​back to the original reference level, thereby accurately reconstructing the actual resource load situation.

[0140] In a specific example, researchers used the previously calculated target range value to perform amplification on the CPU utilization sequence prediction feature of the cloud server. Assuming the target range is 50 and the normalized feature value is 0.6, the amplification process magnifies the feature value to the actual order of magnitude, resulting in a value of 0.6 * 50 = 30. Next, the researchers performed an inverse shift on the amplified feature value based on the minimum sequence prediction feature. Assuming the minimum sequence prediction feature is 20, the inverse shifted feature value is 30 + 20 = 50. Through this method, researchers can accurately obtain resource estimation features that reflect actual CPU utilization changes. Ultimately, the researchers obtained accurate resource estimation features, providing a reliable data foundation for cloud server resource management.

[0141] Sub-step 4033: Take the resource estimation feature components corresponding to each frequency interval in the resource estimation features as reconstruction terms, and perform wavelet reconstruction on the resource estimation features to obtain the resource estimation sequence of the cloud server.

[0142] In some embodiments of this application, the resource estimation feature components corresponding to each frequency interval are used as reconstruction terms to recombine information from different frequency intervals to generate a complete resource estimation sequence. Specifically, the process involves using the resource estimation feature components corresponding to each frequency interval as reconstruction terms and merging these components into a whole through wavelet reconstruction. Wavelet reconstruction can restore the overall shape of the signal, allowing the resource estimation features to reflect the full picture of the resource load after reconstruction. This method can obtain accurate cloud server resource estimation sequences, providing reliable data for resource management and optimization in practical applications.

[0143] In a specific example, researchers reconstructed the cloud server CPU utilization resource estimation features obtained through inverse normalization by frequency range. Assuming the components include approximate component A1 and detail components D1, D2, and D3, the researchers used these components as reconstruction terms and merged them into a single resource estimation sequence using wavelet reconstruction. Specifically, the resource estimation sequence Y was calculated using the formula Y = A1 + D1 + D2 + D3. Through this wavelet reconstruction method, researchers were able to effectively reconstruct the resource load changes of the cloud server and generate an accurate resource estimation sequence. Ultimately, the researchers obtained a complete resource estimation sequence, providing a solid data foundation for subsequent resource management and optimization.

[0144] Step 404: Use the resource estimation sequence as the predicted value of the resource sequence of the cloud server.

[0145] In some embodiments of this application, using the resource estimation sequence as a prediction of the cloud server's resource sequence is to ultimately determine the cloud server's resource load over a future period. Specifically, the process involves using the resource estimation sequence generated in previous steps as input, serving as the predicted value of the cloud server's resource sequence. This process, by using the estimated sequence instead of the actual resource sequence, enables the model to schedule and manage resources based on these predictions, thereby optimizing the cloud server's resource configuration. This method effectively improves the accuracy of predictions and the rationality of resource allocation, ensuring the stable operation of the system.

[0146] In a specific example, researchers used resource estimation sequences generated by a BiLSTM model and a comprehensive evaluation model as predicted resource sequences for cloud servers. These predictions will guide future resource scheduling and allocation for the cloud servers. For instance, assuming the resource estimation sequences are E1, E2, and E3, researchers input these sequences into the system, which will treat them as future resource load data and allocate and optimize resources accordingly. Through this method, researchers can accurately predict the future resource needs of cloud servers, ensuring timely resource allocation when needed, avoiding resource shortages or over-allocation, and ultimately improving the overall efficiency and performance of the system.

[0147] Based on the above process, the prediction performance of the embodiments of this application can be compared with that of different models. The prediction models used for comparison are Support Vector Machine (SVM) and the traditional method of directly using BiLSTM model. The mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) are used as evaluation indicators.

[0148] The formulas for calculating MAE, RMSE, and MAPE are as follows:

[0149]

[0150] In the formula, y predictive y represents the predicted resource load. true This represents the measurement result of resource load, where n represents the number of samplings and i represents the label of the sample.

[0151] Table 1 presents the comparison of prediction performance obtained according to the above comparison.

[0152]

[0153] It can be seen that the method of this application embodiment can significantly reduce prediction error and improve prediction performance.

[0154] Based on the above experiments Figure 7 A comparison chart is provided, showing the prediction results of cloud server resource prediction using BiLSTM under this method and the actual data. It can be seen from the chart that the prediction results obtained by the method of this application embodiment have a high degree of consistency with the original data.

[0155] like Figure 8 The diagram illustrates a complete process for obtaining a resource estimate using the method disclosed in the embodiments of this application:

[0156] S1: Obtain Cloud Server Resource Data: The system obtains resource data from the cloud server, which may include CPU utilization, memory usage, network traffic, etc. Collecting this data provides foundational information for subsequent analysis and prediction.

[0157] S2: Wavelet Decomposition: Wavelet Transform (WT) is used to decompose resource data. Wavelet decomposition can decompose a signal into components of different frequencies, thereby extracting low-frequency and high-frequency features of the data and providing multi-resolution analysis.

[0158] S3: Data Normalization: Normalize the decomposed data (Data Normalization, DN) to map the data to a uniform range (e.g., the [0,1] interval). Normalization can eliminate the magnitude differences between different features and improve the stability of the model.

[0159] S4: Sample Sampling: Sample the normalized data to generate multiple features for training. This step typically involves using techniques such as bootstrap aggregation (bagging) sampling, which involves randomly sampling with replacement from the original dataset. By randomly drawing subsamples from a large sample space, the diversity and representativeness of the data are increased.

[0160] S5: Training a multi-layer bidirectional long short-term memory network model: Use the previously prepared data to train a multi-layer bidirectional long short-term memory (BiLSTM) recurrent neural network model. This model can capture long-term dependencies and short-term fluctuations in time series data, improving prediction accuracy.

[0161] S6: Ensemble Prediction Results: This integrates the results of multiple prediction models, potentially using stacking techniques to combine the strengths of different models. Ensemble learning (EL) can improve overall prediction performance.

[0162] S7: Inverse Normalization: The prediction results are inversely normalized based on the previous normalization parameters, restoring the standardized data to its original magnitude. This provides accurate resource estimation characteristics, facilitating practical applications.

[0163] S8: Wavelet Reconstruction: Wavelet reconstruction is used to merge resource estimation features from different frequency components into a complete resource estimation sequence. Through reconstruction, a sequence reflecting the overall resource load of the cloud server can be obtained, providing reliable data for resource management and optimization.

[0164] Figure 9 The program execution process under the method disclosed in the embodiments of this application is as follows:

[0165] R1: Prepare the training sample set as the basic data for model training.

[0166] R2: Samples are drawn from the training sample set to generate multiple training sets.

[0167] R3: Divide the sampled data into multiple training sets, each of which will be used for independent model training.

[0168] R4: Train the model using each training set to generate multiple trained models.

[0169] R5: Integrates the prediction results of multiple trained models to obtain the final prediction result.

[0170] R6: Outputs the integrated prediction results, completing the entire program execution process.

[0171] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies.

[0172] like Figure 10 As shown in the embodiments of this application, a training device 50 for a data resource prediction model of a cloud server is also disclosed, comprising:

[0173] The sequence module 501 is used to obtain the resource sequence of the cloud server; the resource sequence is used to characterize the change of the cloud server's resource load over time.

[0174] The decomposition module 502 is used to decompose the resource sequence into multiple wavelet sequences to obtain the sequence features of each resource sequence; the sequence features contain multiple feature components, each feature component corresponds to a different frequency range; the feature components are used to characterize the changes in the resource load of the cloud server over time in the frequency range corresponding to the feature component.

[0175] Training module 503 is used to train a bidirectional long short-term memory recurrent neural network based on the sequence characteristics of the resource sequence to obtain a data resource prediction model; the resource prediction model is used to predict the resource load of the cloud server based on the resource sequence to obtain the sequence prediction characteristics of the cloud server.

[0176] Optionally, the decomposition module 502 includes:

[0177] The decomposition submodule is used to decompose the resource sequence into a first resource subsequence and a second resource subsequence; the frequency value in the frequency range corresponding to the first resource subsequence is higher than the frequency value in the frequency range corresponding to the second resource subsequence.

[0178] The iterative submodule is used to record the first resource subsequence obtained by decomposition as a detailed sequence and update the resource sequence with the second resource subsequence when the number of executions of decomposing the resource sequence into the first resource subsequence and the second resource subsequence has not reached the preset decomposition iteration threshold. Then, it returns to the step of decomposing the resource sequence into the first resource subsequence and the second resource subsequence.

[0179] The determination submodule is used to record the second resource subsequence obtained by decomposition as an approximate sequence when the number of executions reaches the decomposition iteration threshold, and to determine the sequence characteristics of the resource sequence data based on all the detailed sequences and the approximate sequence.

[0180] Optionally, the submodules to be determined include:

[0181] The normalization unit is used to determine all the detail sequences and approximate sequences as a resource subsequence of the resource sequence, and to normalize each resource subsequence to obtain the resource normalized subsequence corresponding to the cloud server and each frequency range.

[0182] The feature generation unit is used to obtain sequence features by taking the resource normalized subsequence as the feature component of the sequence feature according to the preset interval order.

[0183] Optionally, training module 503 includes:

[0184] The sampling submodule is used to sample sequence features to obtain multiple sampled features from the cloud server;

[0185] The training submodule is used to train a bidirectional long short-term memory recurrent neural network by obtaining multiple sampled features in order to obtain a trained data resource prediction model.

[0186] Optionally, the sampling submodule includes:

[0187] The slicing unit is used to slice each feature component in the sequence feature according to a preset sampling window threshold to obtain multiple candidate features of the cloud server. The candidate feature components of the candidate features are used to characterize the changes of the cloud server's resource load over time in the frequency interval corresponding to the feature components of the candidate feature components within a preset time window.

[0188] The sampling unit is used to take all the obtained candidate features as the sample space for sampling, randomly extract from the sample space, and take the multiple extracted candidate features as multiple sampling features of the cloud server.

[0189] like Figure 11 As shown in the figure, this application embodiment also discloses a data resource prediction device 60 for a cloud server, including:

[0190] The sampling module 601 is used to obtain the resource sequence of multiple cloud servers according to a preset number of predicted sampling times;

[0191] The prediction module 602 is used to input the resource sequence into the data resource prediction model to obtain multiple sequence prediction features of the cloud server; the data resource prediction model is trained according to the training method of the cloud server data resource prediction model disclosed in the above embodiments.

[0192] The estimation module 603 is used to determine the predicted value of the resource requirements of the cloud server based on the obtained multiple sequence prediction features.

[0193] Optionally, the estimation module 603 includes:

[0194] The estimation submodule is used to estimate the resource requirements of the cloud server by using the obtained multiple sequence prediction features as estimation samples, so as to determine the resource estimation sequence of the cloud server.

[0195] The prediction submodule is used to use the resource estimation sequence as a prediction of the resource sequence of the cloud server.

[0196] Optionally, the estimation submodule includes:

[0197] The normalization estimation unit is used to perform a weighted summation of multiple sequence prediction features to obtain the normalized features of cloud server resource estimation.

[0198] The inverse normalization unit is used to perform inverse normalization processing on the resource estimation normalization features according to the preset normalization features to obtain the resource estimation features of the cloud server; each resource estimation feature component corresponds one-to-one with the frequency interval.

[0199] The wavelet reconstruction unit is used to take the resource estimation feature components corresponding to each frequency interval in the resource estimation features as reconstruction terms and perform wavelet reconstruction on the resource estimation features to obtain the resource estimation sequence of the cloud server.

[0200] Optional, the inverse normalization unit includes:

[0201] The range subunit is used to determine the range of multiple sequence prediction features as the target range value in the process of obtaining multiple sequence prediction features of the cloud server by training the data resource prediction model through the training method.

[0202] The inverse normalization subunit is used to gain the normalized features according to the target range value, and then inversely shift the gained normalized features according to the smallest sequence prediction feature among multiple sequence prediction features to obtain resource estimation features.

[0203] In summary, in this embodiment, by acquiring the resource sequence of the cloud server, the changes in the cloud server resource load over time are comprehensively captured. The resource sequence is then decomposed into multiple wavelet sequences to extract sequence features in different frequency ranges. This allows the model to fully capture the changes in resource load across different frequency ranges, thereby improving prediction accuracy and stability. Furthermore, the BiLSTM model is trained using the decomposed sequence features, enabling it to better capture the temporal characteristics of data resource load and thus improve prediction accuracy. Therefore, the method in this embodiment employs wavelet decomposition to decompose the resource sequence into multiple feature components with different frequency ranges, thereby capturing multi-level information about resource load changes, improving the representativeness of training data and the model's generalization ability, enhancing the temporal modeling capability of the BiLSTM, and solving the problem of insufficient prediction accuracy for cloud server data resources in related technologies.

[0204] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the training method for the data resource prediction model of the cloud server or the various processes of the embodiments of the data resource prediction method for the cloud server, and achieves the same technical effect. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0205] Figure 12 This is a block diagram of an electronic device 700 provided in an embodiment of this application. For example, the electronic device 700 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0206] Reference Figure 12 The electronic device 700 may include one or more of the following components: a processing component 702, a memory 704, a power supply component 706, a multimedia component 708, an audio component 710, an input / output (I / O) interface 712, a sensor component 714, and a communication component 716.

[0207] Processing component 702 typically controls the overall operation of electronic device 700, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 702 may include one or more processors 720 to execute instructions to complete all or part of the steps of the aforementioned cloud server data resource prediction model training method or cloud server data resource prediction method. Furthermore, processing component 702 may include one or more modules to facilitate interaction between processing component 702 and other components. For example, processing component 702 may include a multimedia module to facilitate interaction between multimedia component 708 and processing component 702.

[0208] Memory 704 is used to store various types of data to support the operation of electronic device 700. Examples of this data include instructions for any application or method operating on electronic device 700, contact data, phonebook data, messages, pictures, multimedia, etc. Memory 704 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0209] Power supply component 706 provides power to various components of electronic device 700. Power supply component 706 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 700.

[0210] Multimedia component 708 includes a screen that provides an output interface between electronic device 700 and user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may not only sense the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 708 includes a front-facing camera and / or a rear-facing camera. When electronic device 700 is in an operating mode, such as shooting mode or multimedia mode, the front-facing camera and / or rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0211] Audio component 710 is used to output and / or input audio signals. For example, audio component 710 includes a microphone (MIC) used to receive external audio signals when electronic device 700 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 704 or transmitted via communication component 716. In some embodiments, audio component 710 also includes a speaker for outputting audio signals.

[0212] I / O interface 712 provides an interface between processing component 702 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0213] Sensor assembly 714 includes one or more sensors for providing state assessments of various aspects of electronic device 700. For example, sensor assembly 714 may detect the on / off state of electronic device 700, the relative positioning of components such as the display and keypad of electronic device 700, changes in position of electronic device 700 or a component of electronic device 700, the presence or absence of user contact with electronic device 700, orientation or acceleration / deceleration of electronic device 700, and temperature changes of electronic device 700. Sensor assembly 714 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 714 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0214] Communication component 716 facilitates wired or wireless communication between electronic device 700 and other devices. Electronic device 700 can access wireless networks based on communication standards, such as WiFi, carrier networks (such as 2G, 3G, 4G, or 7G), or combinations thereof. In one exemplary embodiment, communication component 716 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 716 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0215] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to implement the training method for the cloud server data resource prediction model or the cloud server data resource prediction method provided in the embodiments of this application.

[0216] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 704 including instructions, which can be executed by a processor 720 of an electronic device 700 to complete the training method for the cloud server's data resource prediction model or the cloud server's data resource prediction method described above. For example, the non-transitory storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0217] Figure 13 This is a block diagram illustrating an electronic device 800 according to an exemplary embodiment. For example, the electronic device 800 may be provided as a server. (Refer to...) Figure 13 The electronic device 800 includes a processing component 822, which further includes one or more processors, and memory resources represented by memory 832 for storing instructions executable by the processing component 822, such as application programs. The application programs stored in memory 832 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 822 is configured to execute instructions to perform the training method for a cloud server data resource prediction model or the cloud server data resource prediction method provided in the embodiments of this application.

[0218] Electronic device 800 may also include a power supply component 826 configured to perform power management of electronic device 800, a wired or wireless network interface 850 configured to connect electronic device 800 to a network, and an input / output (I / O) interface 858. Electronic device 800 may operate on an operating system stored in memory 832, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0219] This application also provides a computer program product, including a computer program, a training method for a cloud server data resource prediction model, or a cloud server data resource prediction method, implemented when the computer program is executed by a processor.

[0220] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0221] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0222] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0223] It will be readily apparent to those skilled in the art that any combination of the above embodiments is feasible. Therefore, any combination of the above embodiments is an implementation scheme of this application. However, due to space limitations, this specification will not describe them in detail here.

[0224] The training method or cloud server data resource prediction method provided herein is not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. Based on the above description, the required structure for constructing a system with the scheme of this application is obvious. Furthermore, this application is not directed to any particular programming language. It should be understood that the content of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of this application.

[0225] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0226] Similarly, it should be understood that, for the purpose of simplification and aiding understanding of one or more aspects of the application, various features of the application are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the application. However, this approach of disclosure should not be construed as reflecting an intention that the claimed application requires more features than expressly recited in each claim. Rather, as reflected in the claims, the application aspect comprises fewer features than all of the features of the single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the application.

[0227] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0228] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0229] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the cloud server data resource prediction model training method or cloud server data resource prediction method according to the embodiments of this application. This application can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0230] It should be noted that the above embodiments are illustrative of this application and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A training method for a data resource prediction model for a cloud server, characterized in that, include: Obtain the resource sequence of the cloud server; the resource sequence is used to characterize the change of the resource load of the cloud server over time; The sequence features of each resource sequence are obtained by decomposing the resource sequence into multiple wavelet sequences. The sequence feature includes multiple feature components, each of which corresponds to a different frequency range; the feature components are used to characterize the changes in the resource load of the cloud server over time within the frequency range corresponding to the feature component. A bidirectional long short-term memory recurrent neural network is trained based on the sequence characteristics of the resource sequence to obtain a data resource prediction model; the resource prediction model is used to predict the resource load of the cloud server based on the resource sequence to obtain the sequence prediction characteristics of the cloud server. The step of decomposing the resource sequence into multiple wavelet sequences to obtain the sequence features of each resource sequence includes: The resource sequence is decomposed into a first resource subsequence and a second resource subsequence; the frequency value in the frequency range corresponding to the first resource subsequence is higher than the frequency value in the frequency range corresponding to the second resource subsequence. If the number of times the resource sequence is decomposed into a first resource subsequence and a second resource subsequence is less than the preset decomposition iteration threshold, the first resource subsequence obtained by decomposition is recorded as a detail sequence, and the resource sequence is updated with the second resource subsequence. Then, the process returns to the step of decomposing the resource sequence into a first resource subsequence and a second resource subsequence, wherein the detail sequence includes higher frequency components and the updated resource sequence includes lower frequency components. When the number of executions reaches the decomposition iteration threshold, the second resource subsequence obtained by decomposition is recorded as an approximate sequence, and the sequence characteristics of the resource sequence data are determined based on all the detailed sequences and the approximate sequence. Determining the sequence characteristics of the resource sequence data based on all the detailed sequences and the approximate sequences includes: All the detailed sequences and the approximate sequences are respectively determined as a resource subsequence of the resource sequence, and each resource subsequence is normalized to obtain the resource normalized subsequence corresponding to the cloud server and each frequency interval. According to a preset interval order, the resource normalized subsequence is used as the feature component of the sequence feature to obtain the sequence feature.

2. The method as described in claim 1, characterized in that, The step of training a bidirectional long short-term memory recurrent neural network based on the sequence characteristics of the resource sequence to obtain a data resource prediction model includes: The sequence features are sampled to obtain multiple sampled features of the cloud server; By obtaining multiple sampling features, a bidirectional long short-term memory recurrent neural network is trained to obtain a well-trained data resource prediction model.

3. The method as described in claim 2, characterized in that, The sampling of the sequence features to obtain multiple sampled features of the cloud server includes: According to a preset sampling window threshold, each feature component in the sequence feature is sliced ​​into data to obtain multiple candidate features of the cloud server; the candidate feature components of the candidate features are used to characterize the resource load of the cloud server within a preset time window, in the frequency interval corresponding to the feature components of the candidate feature components, and how it changes over time. All the obtained candidate features are used as the sample space for sampling. Randomly sample from the sample space and use the multiple selected candidate features as multiple sampling features of the cloud server.

4. A method for predicting data resources on a cloud server, characterized in that, include: According to the preset number of predicted samplings, obtain the resource sequences of multiple cloud servers; The resource sequences are respectively input into the data resource prediction model to obtain multiple sequence prediction features of the cloud server; The data resource prediction model is trained using the method described in any one of claims 1 to 3; Based on the obtained sequence prediction features, the predicted value of the resource requirement for the cloud server is determined.

5. The method as described in claim 4, characterized in that, The step of determining the predicted value of the resource requirement for the cloud server based on the obtained multiple sequence prediction features includes: The obtained sequence prediction features are used as estimation samples to estimate the resource requirements of the cloud server, so as to determine the resource estimation sequence of the cloud server. The estimated resource sequence is used as the predicted value of the resource sequence of the cloud server.

6. The method as described in claim 5, characterized in that, The step of using the obtained multiple sequence prediction features as estimation samples to estimate the resource requirements of the cloud server, thereby determining the resource estimation sequence of the cloud server, includes: The obtained sequence prediction features are weighted and summed to obtain the resource estimation normalization features of the cloud server. Based on the preset normalization features, the resource estimation normalization features are inversely normalized to obtain the resource estimation features of the cloud server; each component of the resource estimation features corresponds one-to-one with the frequency interval. The resource estimation feature components corresponding to each frequency interval are used as reconstruction terms to perform wavelet reconstruction on the resource estimation features to obtain the resource estimation sequence of the cloud server.

7. An electronic device, characterized in that, include: Processor; memory for storing processor-executable instructions; The processor is configured to execute the instructions to implement the method as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is enabled to perform the method as described in any one of claims 1 to 6.

Images