Performance prediction method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a performance prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: collecting a performance index set of a to-be-monitored target in a historical time period; the performance index set comprises primary indexes and secondary indexes; and predicting a performance prediction result of the to-be-monitored target in the future time period according to the main index and the secondary index. By adopting the method, the performance prediction result can be accurately predicted.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning, and in particular to a performance prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] Traditional performance warning methods usually issue an alarm only when the system is abnormal, which makes it impossible for operation and maintenance personnel to deal with the problem in time, and in serious cases affects the normal operation of the system. In order to improve the stability and reliability of the system, an early warning mechanism that can identify potential performance anomalies in advance is urgently needed.

[0003] The rapid development of deep learning technology has provided new ideas for performance warning. The current mainstream method uses deep learning models, such as Transformer and LSTM, to perform time series prediction on monitoring indicators to identify potential anomalies.

[0004] However, the above methods often rely on a single core indicator for prediction, ignoring the correlation between multiple indicators, resulting in inaccurate prediction results. Summary of the invention

[0005] Based on this, it is necessary to provide a performance prediction method, device, computer equipment, computer-readable storage medium and computer program product that can accurately predict the above technical problems.

[0006] In a first aspect, the present application provides a performance prediction method, the method comprising:

[0007] Collecting a set of performance indicators of the target to be monitored within a historical period; the set of performance indicators includes primary indicators and secondary indicators;

[0008] Based on the main indicators and the secondary indicators, the performance prediction results of the monitored target in a future time period are predicted.

[0009] In one embodiment, predicting the performance prediction result in a future time period according to the primary indicator and the secondary indicator includes:

[0010] Encoding the primary indicators to obtain primary initial features, and encoding the secondary indicators to obtain secondary initial features;

[0011] According to the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are subjected to feature fusion to obtain the target feature; the main feature weight is higher than the secondary feature weight;

[0012] According to the target characteristics, a performance prediction result of the target to be monitored in a future time period is generated.

[0013] In one embodiment, encoding the primary indicator to obtain the primary initial feature, and encoding the secondary indicator to obtain the secondary initial feature, comprises:

[0014] Acquire a feature pattern; the feature pattern is obtained through pre-training;

[0015] According to the feature pattern, the main indicators are mapped to a high-dimensional feature space to obtain the main initial features;

[0016] According to the feature pattern, the secondary indicator is mapped to a high-dimensional feature space to obtain the secondary initial feature.

[0017] In one embodiment, the performance indicator set includes a priority order; and the step of fusing the main initial features and the secondary initial features according to the main feature weights and the secondary feature weights to obtain the target features includes:

[0018] By linearly changing the parameters corresponding to the network, the reference main feature weight and the reference secondary feature weight are obtained;

[0019] Adjusting the reference main feature weight and the reference secondary feature weight according to the priority order to obtain the main feature weight and the secondary feature weight;

[0020] Based on the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are weighted to obtain the target feature.

[0021] In one embodiment, generating a performance prediction result in a future time period according to the target feature includes:

[0022] Performing a linear transformation on the target feature to obtain a linear transformation result;

[0023] Activating the linear transformation result through an activation function to obtain an activation feature;

[0024] According to the activation feature, a performance prediction result of the target to be monitored in a future time period is obtained.

[0025] In one embodiment, the performance prediction result of the monitored target in a future time period is predicted based on the primary indicator and the secondary indicator by a prediction model; the training process of the prediction model includes:

[0026] Acquire a performance indicator sample; the performance indicator sample carries a label;

[0027] Inputting the performance indicator samples into the initial model for training to obtain predicted values;

[0028] A target loss function is obtained based on the predicted value and the label, and the parameters of the initial model are adjusted according to the target loss function until the training is completed to obtain the prediction model.

[0029] In a second aspect, the present application provides a performance prediction device, the device comprising:

[0030] An acquisition module is used to collect a set of performance indicators of the target to be monitored within a historical period of time; the set of performance indicators includes primary indicators and secondary indicators;

[0031] The prediction module is used to predict the performance prediction results of the monitored target in a future time period based on the main indicators and the secondary indicators.

[0032] In a third aspect, the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0033] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0034] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the steps of the above method when executed by a processor.

[0035] The above-mentioned performance prediction method, apparatus, computer device, computer-readable storage medium and computer program product, in combination with the primary and secondary indicators of the target to be monitored, can accurately predict the performance indicators of the target to be monitored in the future time period. This is because the primary indicators reflect the core state of the target to be monitored, while the secondary indicators provide supplementary information of the target to be monitored. By combining these indicators, performance changes can be captured more comprehensively, thereby improving the accuracy of performance prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0037] FIG1 is a schematic diagram of a flow chart of a performance prediction method in one embodiment;

[0038] FIG2 is a schematic diagram of weight generation in one embodiment;

[0039] FIG3 is a schematic diagram of a prediction model structure in one embodiment;

[0040] FIG4 is a block diagram of a performance prediction device in one embodiment;

[0041] FIG. 5 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0043] In one embodiment, as shown in FIG1 , a performance prediction result method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0044] Step 102, collecting a set of performance indicators within a time period; the set of performance indicators includes primary indicators and secondary indicators.

[0045] Among them, the performance indicator set is a set of multiple performance indicators generated by the operation of the monitored target during the historical time period; the monitored target refers to the target whose performance status and workload need to be monitored, which can be a database system, application server, distributed storage node, container cluster, virtual machine instance, etc.

[0046] Among them, the main indicators are performance indicators that have a direct impact on the warning problem, the secondary indicators are performance indicators that have an indirect impact on the problem, and the secondary indicator values ​​are performance metrics introduced to support and supplement the main indicators.

[0047] For example, if the system resource usage is to be warned, the CPU usage rate can be used as the main indicator, which can directly reflect the load and resource usage of the target to be monitored. The query volume (QPS), transaction volume (TPS), number of active sessions and number of slow SQL can be set as secondary indicators. These secondary indicators can help the system fully understand the operating status of the target to be monitored, so as to enhance the prediction accuracy of the target to be monitored.

[0048] Optionally, historical performance indicators can be collected at preset time intervals, such as collecting data every 5 minutes, to ensure the timeliness and representativeness of the data. The collection interval can be flexibly set according to the fluctuation frequency of the system load and business needs. A shorter collection interval can capture more details, but will increase the burden of data storage and calculation; a longer interval can reduce the system load, but may not be able to respond to rapidly changing load conditions in a timely manner.

[0049] Optionally, each performance indicator in the performance indicator set is sorted according to priority, wherein important indicators have a higher priority than secondary indicators.

[0050] Optionally, priority refers to the relative importance of different performance indicators in the monitoring system, which is used to help the model pay more attention to indicators that have a greater impact on the overall performance of the system during analysis and prediction. Priority can be obtained based on business needs, system framework characteristics, and expert experience. For example, in database performance monitoring, core indicators such as CPU usage and memory usage have a higher priority because they directly affect the response time and stability of the system, while secondary indicators such as QPS, TPS, and the number of slow SQLs are used as auxiliary information to provide more comprehensive context support.

[0051] Optionally, the performance indicator set may include one or more important indicators. For example, in a large distributed database environment, important indicators may include CPU usage and memory usage. Changes in these indicators directly indicate whether the node is overloaded or has performance bottlenecks. Other secondary indicators serve as supporting information to provide a broader perspective for analysis. According to different business needs, the performance indicator set can be flexibly configured so that important indicators are given priority during data collection, analysis, and model prediction.

[0052] Step 104: predict the performance prediction result of the target to be monitored in a future time period based on the main indicators and the secondary indicators.

[0053] Optionally, the system performance in the future is predicted based on the primary and secondary indicators obtained from the historical performance indicator set. Performance prediction is a time series analysis that predicts future operating status by modeling and analyzing past performance data.

[0054] Optionally, the data of the primary and secondary indicators can be processed before forecasting. The first step of data processing is to divide the time series data into multiple subsequences, each of which contains historical data of multiple time steps. The purpose of this is to use the data of multiple time steps as the input of the forecasting model so that the forecasting model can learn the correlation between different time steps from the time series.

[0055] Optionally, when making predictions, the primary metric is given a larger weight, while the secondary metric is given a smaller weight. The weights of the secondary metrics and the weights of the primary metrics are generated based on the priority in the performance metric results. The prediction model strengthens the influence of the primary metric through weighted processing or adaptive fusion layers to ensure that the prediction model is more sensitive to the core indicators of system load. Secondary metrics are used as auxiliary features to provide additional contextual information to the model. For example, if some nodes have a long response time but the CPU and memory load are not high, secondary metrics in this case (such as query frequency, network latency, etc.) can help the model more comprehensively evaluate the operating status of the system. Next, the model will perform feature extraction and pattern recognition based on these multi-dimensional features to find trends that may affect the future load of the system.

[0056] Performance prediction results are usually a collection of performance states within a certain period of time in the future. These results may include the numerical ranges of major indicators such as CPU usage, memory usage, response time, etc. every minute or hour in the future.

[0057] Optionally, to further improve the accuracy of performance prediction results, confidence intervals can be added to the prediction results, which can quantify the credibility of the prediction. For some indicators with large changes, the confidence intervals can be expanded or contracted by adjusting the model parameters to make the prediction results more stable and reliable.

[0058] Optionally, after the prediction is completed, the system can make dynamic adjustments based on the prediction results, such as optimizing resource allocation, adjusting load strategies, etc., to help the system maintain smooth operation during high load or potential abnormalities.

[0059] In the above embodiment, the performance indicators of the target to be monitored in the future time period can be accurately predicted by combining the main indicators and the secondary indicators. This is because the main indicators reflect the core status of the target to be monitored, while the secondary indicators provide supplementary information of the target to be monitored. By combining these indicators, performance changes can be captured more comprehensively, thereby improving the accuracy of performance prediction results.

[0060] Furthermore, in one embodiment, the above-mentioned prediction of the performance prediction results in the future time period based on the main indicators and the secondary indicators includes: encoding the main indicators to obtain the main initial features; encoding the secondary indicators into secondary initial features; performing feature fusion on the main initial features and the secondary initial features according to the main feature weights and the secondary feature weights to obtain the target features; the main feature weight is higher than the secondary feature weight; and generating the performance prediction results in the future time period according to the target features.

[0061] First, the main indicators and secondary indicators are encoded respectively to obtain the main initial features and secondary initial features. Encoding refers to the process of converting the original, possibly high-dimensional and noisy time series data or original indicator values ​​into a feature vector that is more operational and can reflect the underlying pattern through a conversion mechanism.

[0062] Furthermore, the above encoding of the main indicators to obtain the main initial features and encoding of the secondary indicators as secondary initial features include: obtaining feature patterns; the feature patterns are obtained by pre-training; according to the feature patterns, the main indicators are mapped to the high-dimensional feature space to obtain the main initial features; according to the feature patterns, the secondary indicators are mapped to the high-dimensional feature space to obtain the secondary initial features.

[0063] Optionally, the encoder layer can be pre-trained to obtain feature patterns. The pre-training process uses large-scale historical data to learn the parameters of the encoder layer, so that it can effectively map different performance indicators to a high-dimensional feature space. Through pre-training, the encoder can capture the potential relationships and regularities between these performance indicators, thereby learning a common feature pattern. The feature pattern reflects the feature distribution learned by the encoder layer when processing the input data. By pre-training the encoder, the subsequent model can still extract effective feature information from the existing historical data when facing small sample data.

[0064] Among them, the pre-trained encoder has learned some common feature representations through a large amount of historical data. These feature representations are not only applicable to the current task, but can also be transferred when facing new data. For example, when there is a small amount of new data, that is, when there is a small sample problem, directly training a brand new model with this data may lead to overfitting, because the model may overfit the noise and anomalies in these small data sets. However, if the encoder is trained with a large amount of historical data in advance, then when facing new data, the pre-trained model will learn some common rules from the historical data, such as the change pattern of performance indicators, the correlation between different indicators, etc. Therefore, even when facing small sample data, the model can still use the knowledge it has learned to extract effective features from the new data and make reasonable predictions.

[0065] Optionally, according to the rules of the feature pattern, the main indicators are mapped to a high-dimensional feature space to obtain the main initial features. "Mapping" here means mapping the original indicator value or time series data to a new space through some mathematical transformation (such as matrix multiplication, nonlinear activation function, etc.), so that the representation of the data is more suitable for subsequent prediction tasks. The same process is also applied to the secondary indicators, which are mapped to the high-dimensional feature space through the feature pattern to obtain the secondary initial features.

[0066] Next, the model needs to fuse the primary initial features and secondary initial features according to the primary feature weights and secondary feature weights to obtain the target feature. The purpose of feature fusion is to combine the secondary indicators with the primary indicators to generate a target feature for subsequent prediction tasks. This is also called weighting based on the importance of the indicator, that is, the weight.

[0067] Furthermore, the above-mentioned performance indicator set includes a priority order; according to the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are feature fused to obtain the target feature, including: obtaining the reference main feature weight and the reference secondary feature weight through the parameters corresponding to the linear change network; adjusting the reference main feature weight and the reference secondary feature weight according to the priority order to obtain the main feature weight and the secondary feature weight; based on the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are weighted to obtain the target feature.

[0068] Specifically, the weight of each feature is calculated through a linear transformation network. In this network, the model is trained to obtain a set of parameters that map the input primary and secondary features to weighted output features through linear transformation. This process can be done in the following way: First, the network automatically learns a reference weight based on the existing training data, which corresponds to the relative importance of the primary and secondary features. Then, these reference weights are further adjusted according to the model requirements to ensure that they are consistent with the priority of each feature.

[0069] Optionally, in conjunction with FIG2 , FIG2 is a schematic diagram of weight generation in an embodiment. A priority vector may be used, such as [1, 2, 3, 4, 5], indicating that the CPU usage is 1, the QPS is 2, and so on. At this time, the reference weight will be dynamically adjusted according to the priority vector to obtain the main feature weight and the secondary feature weight.

[0070] Furthermore, optionally, the weight can be adjusted using the inverse of the priority vector (or exponential function), that is, adjusted weight = reference weight × inverse of priority.

[0071] For example, CPU usage: reference weight 0.2 × 1 / 1 = 0.2; QPS: reference weight 0.3 × 1 / 2 = 0.15.

[0072] After obtaining the adjusted primary and secondary feature weights, the model weights the primary and secondary initial features according to these weights. These weighted features are passed as input to the subsequent prediction step.

[0073] Optionally, weighting can be performed in a variety of ways, for example, using weighted sums, weighted averages, or using neural networks for weighted calculations in more complex models. The core idea of ​​weighting is to give more weight to important features, thereby ensuring that they contribute more to the prediction results.

[0074] Specifically, the weighted fusion feature is the target feature, which includes the weighted information of all primary and secondary indicators and can be used for the next performance prediction or anomaly detection task. The generation of target features is a process that combines weight adjustment and feature weighting, and its purpose is to improve the prediction accuracy and stability of the model and ensure that the influence of each feature can be properly reflected in the overall prediction.

[0075] Finally, the fused target features will be used as the input of the prediction model and enter the next prediction stage. Based on these fused features, the model uses the patterns learned during the training process to predict the performance in the future time period. This prediction process is based on time series analysis, which predicts the performance prediction results of the system in the future time period through the trends and rules of historical data.

[0076] Furthermore, the above-mentioned generating performance prediction results for a future time period based on the target features includes: performing a linear transformation on the target features to obtain a linear transformation result; activating the linear transformation result through an activation function to obtain an activated feature; and obtaining a performance prediction result for a future time period based on the activated feature.

[0077] The target features are obtained by fusing the primary features and the secondary features, and already contain high-dimensional information related to the target task. These target features may be expressed as a vector or matrix, representing the weighted fusion result of all performance indicators at the current moment. Through linear transformation, the target features will be mapped to a new space to obtain a new representation with different semantic features. The role of linear transformation is to compress or map the high-dimensional representation of the target features to a new space, so that the model can better learn patterns related to future performance prediction through this new representation.

[0078] Next, an activation function is applied to the linear transformation result Z. The role of the activation function is to introduce nonlinear factors, thereby enhancing the expressive power of the model and allowing the model to capture more complex patterns. Through the activation function, the linear transformation result Z is mapped to a nonlinear space. In this way, the model can learn nonlinear relationships between more complex features and further improve prediction accuracy.

[0079] Finally, the fully connected layer performs a linear change on the activated features and calculates them with the weights and biases of the model to obtain the performance prediction results of the monitored target in the future time period.

[0080] Through the above steps, the model can combine the characteristic information of multiple performance indicators to accurately predict the performance prediction results of the monitored target in the future time period.

[0081] In one embodiment, the performance prediction results in the future time period are predicted based on the main indicators and the secondary indicators through a prediction model; the training process of the prediction model includes: obtaining performance indicator samples; the performance indicator samples carry labels; the performance indicator samples are input into the initial model for training to obtain predicted values; the target loss function is obtained based on the predicted values ​​and the labels, and the parameters of the initial model are adjusted according to the target loss function until the training is completed to obtain the prediction model.

[0082] First, obtain performance indicator samples. Performance indicator samples refer to the performance indicator data of the target to be monitored within a certain historical period. These indicators may include CPU usage, QPS, TPS, memory usage, etc. Each sample data contains not only the actual values ​​of these indicators, but also carries label data. Label data usually indicates whether performance anomalies or warning events occur within the period.

[0083] Next, the performance indicator samples are input into the initial model for training. The initial model is usually a deep learning-based neural network model, such as LSTM or Transformer, which can process time series data and generate predictions for future performance. By inputting performance indicator samples, the model begins to learn patterns and regularities in historical data.

[0084] The model then calculates the target loss function by comparing the predicted value with the true label. The target loss function measures the deviation of the model's prediction results from the actual results, such as using the mean squared error (MSE) or cross entropy loss function.

[0085] Finally, based on the feedback from the loss function, the model parameters, such as weights and biases, are adjusted through the back-propagation algorithm. Through multiple iterations of training, the model parameters are gradually optimized until the loss function reaches the preset stop criteria and the training process is completed. At this point, the prediction model has a strong prediction capability and can be used to predict future performance indicators.

[0086] In the above embodiment, a prediction model for predicting the target to be monitored in a future time period can be obtained through training.

[0087] In an exemplary embodiment, the target to be monitored is a database, and the performance prediction method includes:

[0088] 1. Obtain the performance monitoring data of each indicator of the database in the current time period, that is, obtain the performance indicator set.

[0089] 2. Set priorities for each performance indicator in the performance indicator set to obtain primary and secondary indicators.

[0090] 3. Input the main index and the secondary index into the prediction model. For example, referring to FIG3 , FIG3 is a schematic diagram of the prediction model structure in one embodiment.

[0091] 3.1, encode the main features and secondary features. Since the pre-trained encoder has learned the general feature representation, that is, the feature pattern mentioned in the above embodiment. It has better robustness against small samples of abnormal data and noise interference of performance indicators. The main function of this layer is to encode the collected original feature X into a general feature representation, that is, H=Encoder(X), H represents the high-dimensional feature representation obtained by the encoder, X={x1,x2,...,x n} represent n performance monitoring indicators in sequence, where x n is a t time series vector {x n 1 ,x n 1 ,..,x n t}, indicating the indicator time series of the nth indicator.

[0092] 3.2, according to the main feature weights and the secondary feature weights, the main initial features and the secondary initial features are fused to obtain the target features. As shown in Figure 2, this layer defines a weight network and a priority vector. The weight network is a linear transformation Ln, whose parameters are automatically adjusted during the fine-tuning process. The multi-index fusion weight Ln (H) is adaptive to the business characteristics during the fine-tuning process. At the same time, in order to introduce prior knowledge into the model, the model is prompted to consider the priority relationship of the indicators in the adaptive weight. Since the indicators are arranged in priority order when they are input, the vector position of each indicator implies its priority relationship. Therefore, the position index of each vector is used for hard coding v p =[1,2,..,n]. The final adaptive fusion weight expression is:

[0093] W = softmax(Ln(H)+vp) Formula (1)

[0094] On the one hand, the weight learns the association between the indicator vectors, and on the other hand, it integrates the prior primary and secondary priorities. Softmax converts the result into a weighted representation. The final calculation of the adaptive fusion layer is:

[0095] AdaFusion(H)=W*H formula (2)

[0096] 3.3, based on the target features, generate the performance prediction results of the database in the future time period. The fully connected layer consists of a linear transformation and an activation function stack, which is used to regularize the intermediate representation of the features and convert it into the warning state Y at the next u moments. The Yu value at time u is 0 or 1, and 1 means that time u is marked as a performance prediction result.

[0097] In the above embodiment, a pre-trained time series encoder is used for modeling. Different from the conventional training from scratch by collecting monitoring data, it reduces the computing cost and improves the effectiveness of the early warning method based on the prediction of monitoring indicators. It also fuses various monitoring indicators through adaptive weights. It is only necessary to fine-tune the adaptive fusion layer in the specific business scenario, thereby enhancing the accuracy of performance warnings and the adaptability and stability to different business scenarios.

[0098] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0099] Based on the same inventive concept, the embodiment of the present application also provides a performance prediction device for implementing the above-mentioned performance prediction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more performance prediction device embodiments provided below can refer to the limitations on the performance prediction method above, and will not be repeated here.

[0100] In an exemplary embodiment, as shown in FIG4 , a performance prediction device is provided, including: an acquisition module 410 and a prediction module 420, wherein:

[0101] The acquisition module 410 is used to collect a set of performance indicators of the target to be monitored within a historical period of time; the set of performance indicators includes primary indicators and secondary indicators.

[0102] The prediction module 420 is used to predict the performance prediction results of the monitored target in a future time period based on the main indicators and the secondary indicators.

[0103] In one embodiment, the prediction module 420 includes:

[0104] The encoding unit is used to encode the main indicators to obtain the main initial features, and to encode the secondary indicators to obtain the secondary initial features.

[0105] The feature fusion unit is used to fuse the main initial features and the secondary initial features according to the main feature weight and the secondary feature weight to obtain the target feature; the main feature weight is higher than the secondary feature weight.

[0106] The feature prediction unit is used to generate performance prediction results of the target to be monitored in a future time period based on the target features.

[0107] In one embodiment, the encoding unit includes:

[0108] The pattern acquisition subunit is used to acquire feature patterns; the feature patterns are obtained through pre-training.

[0109] The first mapping subunit is used to map the main indicators to the high-dimensional feature space according to the feature pattern to obtain the main initial features.

[0110] The second mapping subunit is used to map the secondary index to the high-dimensional feature space according to the feature pattern to obtain the secondary initial feature.

[0111] In one embodiment, the feature fusion unit includes:

[0112] The reference feature acquisition subunit is used to obtain the reference main feature weight and the reference secondary feature weight by linearly changing the corresponding parameters of the network.

[0113] The adjustment subunit is used to adjust the reference main feature weight and the reference secondary feature weight according to the priority order to obtain the main feature weight and the secondary feature weight.

[0114] The weighting subunit is used to weight the main initial features and the secondary initial features based on the main feature weight and the secondary feature weight to obtain the target feature.

[0115] In one embodiment, the feature prediction unit includes:

[0116] The transformation subunit is used to perform linear transformation on the target feature to obtain a linear transformation result.

[0117] The activation subunit is used to activate the linear transformation result through the activation function to obtain the activation feature.

[0118] The result generation subunit is used to obtain the performance prediction result of the target to be monitored in the future time period according to the activated features.

[0119] In one embodiment, the apparatus comprises a training module, the training module comprising:

[0120] The sample acquisition unit is used to acquire performance indicator samples; the performance indicator samples carry labels.

[0121] The model prediction unit is used to input the performance indicator samples into the initial model for training to obtain the predicted value.

[0122] The model training unit is used to obtain the target loss function based on the predicted value and the label, and adjust the parameters of the initial model according to the target loss function until the training is completed to obtain the prediction model.

[0123] Each module in the above performance prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0124] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG5. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store a set of performance indicators. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a performance prediction method is implemented.

[0125] Those skilled in the art will understand that the structure shown in FIG. 5 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0126] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps of the method in any one of the above embodiments when executing the computer program.

[0127] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in any one of the above embodiments are implemented.

[0128] In one embodiment, a computer program product is provided, comprising a computer program, which implements the steps of the method in any one of the above embodiments when executed by a processor.

[0129] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0130] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0131] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A performance prediction method, characterized in that: The method comprises: Collecting a set of performance indicators of the target to be monitored within a historical period; the set of performance indicators includes primary indicators and secondary indicators; Based on the main indicators and the secondary indicators, the performance prediction results of the monitored target in a future time period are predicted.

2. The method according to claim 1, characterized in that The predicting of the performance prediction results in the future time period according to the main indicator and the secondary indicator includes: Encoding the primary indicators to obtain primary initial features, and encoding the secondary indicators to obtain secondary initial features; According to the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are subjected to feature fusion to obtain the target feature; the main feature weight is higher than the secondary feature weight; According to the target characteristics, a performance prediction result of the target to be monitored in a future time period is generated.

3. The method according to claim 2, characterized in that The encoding of the primary indicator to obtain the primary initial feature and the encoding of the secondary indicator to obtain the secondary initial feature include: Acquire a feature pattern; the feature pattern is obtained through pre-training; According to the feature pattern, the main indicators are mapped to a high-dimensional feature space to obtain the main initial features; According to the feature pattern, the secondary indicator is mapped to a high-dimensional feature space to obtain the secondary initial feature.

4. The method according to claim 2, characterized in that: The performance indicator set includes a priority order; the feature fusion of the main initial features and the secondary initial features according to the main feature weights and the secondary feature weights to obtain the target features includes: By linearly changing the parameters corresponding to the network, the reference main feature weight and the reference secondary feature weight are obtained; Adjusting the reference main feature weight and the reference secondary feature weight according to the priority order to obtain the main feature weight and the secondary feature weight; Based on the main feature weight and the secondary feature weight, the main initial feature and the secondary initial feature are weighted to obtain the target feature.

5. The method according to claim 2, characterized in that: Generating a performance prediction result in a future time period according to the target feature includes: Performing a linear transformation on the target feature to obtain a linear transformation result; Activating the linear transformation result through an activation function to obtain an activation feature; According to the activation feature, a performance prediction result of the target to be monitored in a future time period is obtained.

6. The method according to claim 1, characterized in that The prediction result of the performance of the target to be monitored in a future time period is predicted based on the main indicator and the secondary indicator by a prediction model; The training process of the prediction model includes: Acquire a performance indicator sample; the performance indicator sample carries a label; Inputting the performance indicator samples into the initial model for training to obtain predicted values; A target loss function is obtained based on the predicted value and the label, and the parameters of the initial model are adjusted according to the target loss function until the training is completed to obtain the prediction model.

7. A performance prediction device, characterized in that: The device comprises: An acquisition module is used to collect a set of performance indicators of the target to be monitored within a historical period of time; the set of performance indicators includes primary indicators and secondary indicators; The prediction module is used to predict the performance prediction results of the monitored target in a future time period based on the main indicators and the secondary indicators.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

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

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

Images