A fault prediction method, apparatus, computing device, and storage medium

Abstract

This application provides a fault prediction method, apparatus, computing device, and storage medium. It can filter target base model pools based on the numerical change patterns corresponding to different performance parameters, dynamically sort the base models in the target base model pools, select a set number of target base models with the highest prediction accuracy, use the target base models to perform numerical prediction to obtain the numerical prediction results, and then determine the corresponding fault prediction results. This method avoids the impact of changes in data distribution on the fault prediction process. Different base model pools have different processing capabilities for different data change patterns. For different data distributions, i.e., different data change pattern types, a base model with stronger processing capabilities can be selected to handle the corresponding data change pattern type, ensuring prediction accuracy, avoiding fault occurrences and data loss, and improving the user experience.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a fault prediction method, apparatus, computing device, and storage medium. Background Technology

[0002] With the development of the supercomputing internet, in order to improve computing power, various enterprises and organizations are connecting to the supercomputing internet in their respective businesses, and connecting the supercomputing internet to clusters and nodes on a large scale. In the process of using the supercomputing internet, it is necessary to quickly identify and repair faulty nodes to prevent data loss and improve the user experience.

[0003] Existing node failure prediction technologies mainly rely on monitoring and analyzing the performance parameters of physical machines, comparing the obtained performance parameter values ​​with static thresholds, and considering node failure as occurring once the data exceeds the threshold. However, the accuracy of prediction is affected when the data distribution changes, making it prone to failure and data loss, which impacts the user experience. Summary of the Invention

[0004] To address the problems in the prior art, this application provides a fault prediction method, apparatus, computing device, and storage medium, which can predict whether a node will fail under different data distributions using different methods, and can ensure stability in the fault prediction process, avoid data loss, and guarantee user experience.

[0005] In a first aspect, embodiments of this application provide a fault prediction method, including:

[0006] Based on the generation time of multiple values ​​of the same performance parameter and the magnitude relationship of the multiple values, the change pattern type of the multiple values ​​is determined, and the target base model pool corresponding to the determined change pattern type is selected from multiple base model pools.

[0007] The first part of the multiple values ​​is input into each target base model in the target base model pool, and the prediction accuracy of each target base model is determined based on the second part of the multiple values ​​and the values ​​output by each target base model; wherein the generation time of the first part of the values ​​is earlier than the generation time of the second part of the values.

[0008] Based on the prediction accuracy of each target base model, a set number of target base models are selected;

[0009] The second part of the numerical values ​​are input into the set number of target base models to obtain at least one numerical prediction result, and the fault prediction result is determined based on the at least one numerical prediction result.

[0010] This method avoids the impact of changes in data distribution on the fault prediction process. Different base model pools can be used to process different data distributions, i.e., different types of data change patterns. The base models in different base model pools have different processing capabilities for different data change patterns. Selecting a base model with stronger processing capabilities to process the corresponding data change pattern type can ensure the accuracy of prediction, avoid faults and data loss, and improve the user experience.

[0011] In one possible embodiment, before determining the type of variation pattern of the multiple values ​​based on the generation time of the acquired values ​​of the same performance parameter and the magnitude relationship of the multiple values, the method further includes:

[0012] Periodically obtain the values ​​of performance parameters;

[0013] For each performance parameter value obtained, the relationship between the performance parameter value and the value range is determined; the value range is determined based on the obtained performance parameter values.

[0014] If the value of the performance parameter is not within the range, the target value determined based on the range will be used as the value of the performance parameter.

[0015] During the process of obtaining performance parameter values, the obtained performance parameters can be preprocessed to avoid the impact on the subsequent fault prediction process due to the failure or error of obtaining some performance parameters. This can improve the accuracy of the fault prediction process to a certain extent.

[0016] In one possible embodiment, the value range is determined based on the obtained performance parameter values ​​in the following manner:

[0017] For each N performance parameter values ​​obtained, the value range is determined based on some or all of the obtained N performance parameter values, a preset upward floating range, and a preset downward floating range; where N is a positive integer.

[0018] By determining the corresponding value range before obtaining the numerical values ​​of performance parameters using the above method, and by using different numbers of performance parameters to determine the value range, the value range can be determined more accurately.

[0019] In one possible embodiment, the base models in the base model pool are obtained in the following manner:

[0020] The multiple sets of training data are classified according to each type of change pattern, so that each type of change pattern corresponds to at least one set of training data; each set of training data includes data to be processed and validation data;

[0021] For any base model, at least one set of training data corresponding to each of the aforementioned change pattern types is input into the base model, and the fitting result corresponding to each of the aforementioned change pattern types is determined based on the output training results and the validation data.

[0022] Based on the multiple fitting results, the corresponding change pattern type of any base model;

[0023] At least one base model corresponding to each of the aforementioned change pattern types is saved to the base model pool corresponding to the aforementioned change pattern type.

[0024] By classifying multiple base models in the above manner, the performance parameters of each base model are determined to be more suitable for handling different types of change patterns. A corresponding base model pool is set up for each type of change pattern to store the corresponding base model. By classifying the base models, it is possible to avoid comparing the performance parameter handling capabilities of each base model during the fault prediction process, thereby reducing the time required for fault prediction and improving the efficiency of the fault prediction process.

[0025] In one possible embodiment, determining the fault prediction result based on the at least one numerical prediction result includes:

[0026] If the number of numerical prediction results obtained is one, the fault prediction result is determined based on the change pattern type corresponding to the final numerical prediction result and the numerical prediction result, wherein the final numerical prediction result is the numerical prediction result.

[0027] If the number of numerical prediction results obtained is greater than one, then the multiple prediction results are linearly fitted to obtain the final numerical prediction result, and the fault prediction result is determined based on the final numerical prediction result and the corresponding change pattern type.

[0028] Different processing methods are determined for different numbers of numerical prediction results. If there are multiple numerical prediction results, the most accurate numerical prediction result can be determined based on multiple numerical prediction results through linear fitting, which improves the efficiency and accuracy of the fault prediction process.

[0029] In one possible embodiment, determining the fault prediction result based on the final numerical prediction result and the corresponding change pattern type includes:

[0030] If the variation pattern of the multiple values ​​is of the first type, and the difference between the final value prediction result and the maximum value among the multiple performance parameters is greater than the first threshold, then the fault prediction result is determined to be a fault occurrence.

[0031] If the pattern of change of the multiple values ​​is of the second type, and the final value prediction result is greater than the second threshold, then the fault prediction result is determined to be a fault occurrence.

[0032] If the variation pattern of the multiple values ​​is of type three, and the final value prediction result is not within the target confidence interval, then the fault prediction result is determined to be a fault occurrence; the target confidence interval is determined based on the maximum and minimum values ​​among the multiple values ​​of the same performance parameter.

[0033] Determining different fault prediction methods based on different patterns of numerical change allows for a more targeted determination of the optimal fault prediction method for each pattern, thereby improving the efficiency and accuracy of the fault prediction process.

[0034] In one possible embodiment, after determining the fault prediction result based on the at least one numerical prediction result, the method further includes:

[0035] If the fault prediction result indicates that a fault has occurred, the user is prompted with fault information. The fault information includes at least one of the following: the performance parameter corresponding to the fault, the type of change pattern of the performance parameter corresponding to the fault, the value of the performance parameter corresponding to the fault, and the fault repair measures.

[0036] After a fault is identified, the system can display fault information to the user, allowing them to repair the fault based on the repair measures contained in the fault information. Alternatively, the system can determine the most suitable repair measures based on other information, thereby improving the efficiency of the fault repair process.

[0037] Secondly, embodiments of this application provide a fault prediction device, comprising:

[0038] The target base model pool determination unit is used to determine the change pattern type of the multiple values ​​based on the generation time of multiple values ​​of the same performance parameter and the size relationship of the multiple values, and select the target base model pool corresponding to the determined change pattern type from multiple base model pools.

[0039] The prediction accuracy determination unit is used to input the first part of the plurality of values ​​into each target base model in the target base model pool, and determine the prediction accuracy of each target base model based on the second part of the plurality of values ​​and the values ​​output by each target base model; wherein the generation time of the first part of the values ​​is earlier than the generation time of the second part of the values.

[0040] The target basis model determination unit is used to select a set number of target basis models based on the prediction accuracy of each target basis model.

[0041] The fault prediction result determination unit is used to input the second part of the numerical values ​​into the set number of target base models respectively to obtain at least one numerical prediction result, and determine the fault prediction result based on the at least one numerical prediction result.

[0042] Thirdly, embodiments of this application provide a computing device, including:

[0043] Memory, used to store program instructions;

[0044] A processor is configured to invoke program instructions stored in the memory and execute the steps included in the method described in the first aspect according to the obtained program instructions.

[0045] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0046] This application provides a method for filtering target base model pools based on the numerical variation patterns of different performance parameters, dynamically sorting the base models in the target base model pools, selecting a set number of target base models with the highest prediction accuracy, using the target base models to perform numerical prediction to obtain numerical prediction results, and then using the numerical prediction results to determine the corresponding fault prediction results. This method can avoid the impact of changes in data distribution on the fault prediction process. The base models in different base model pools have different processing capabilities for different data variation patterns. For different data distributions, that is, different data variation pattern types, a base model with stronger processing capabilities can be selected to process the corresponding data variation pattern type, which can ensure the accuracy of prediction, avoid fault occurrence and data loss, and improve the user experience. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram illustrating an application scenario of a fault prediction method provided in an embodiment of this application.

[0049] Figure 2 A flowchart illustrating a fault prediction method provided in an embodiment of this application;

[0050] Figure 3 This is a schematic diagram illustrating a specific process of a fault prediction method provided in an embodiment of this application;

[0051] Figure 4 This is a schematic diagram of the structure of a fault prediction device provided in an embodiment of this application;

[0052] Figure 5 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0054] In the description of the embodiments of this application, unless otherwise stated, "and" means "or", for example, A / B can mean A or B; "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone.

[0055] Specifically, in the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "connection" and "linkage" used in this application, unless otherwise specified, include both direct and indirect connections (linkages).

[0056] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0057] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0058] In one possible embodiment, Figure 1 This illustration shows an application scenario diagram of a fault prediction method provided in an embodiment of this application. See also: Figure 1 As shown, the computing device cluster 100 includes multiple computing devices, such as computing device 110, computing device 120, computing device 130, etc. The fault prediction method provided in this application embodiment can be applied to each computing device in the computing device cluster 100. Each computing device in the computing device cluster 100 can be a graphics processing unit (GPU) server, a terminal device, or other devices capable of data processing and computation; this application does not limit the specific device.

[0059] Figure 1 The application scenario described above is merely an example of one application scenario for implementing the embodiments of this application, and the embodiments of this application are not limited to the above. Figure 1 The application scenarios described above are illustrated below with reference to the accompanying drawings. It should be noted that the application scenarios described above are merely illustrative for the purpose of understanding the spirit and principles of this application, and the implementation methods of this application are not limited in any way.

[0060] Figure 2This document illustrates a flowchart of a fault prediction method provided in an embodiment of this application. Figure 2 As shown, the method may include the following steps:

[0061] Step S201: Based on the generation time of multiple values ​​of the same performance parameter and the size relationship of the multiple values, determine the change pattern type of the multiple values, and select the target base model pool corresponding to the determined change pattern type from multiple base model pools.

[0062] In one possible embodiment, before executing step S201, the values ​​of performance parameters can be acquired periodically. The period can be once per second or once every 0.1 seconds; this application does not limit the acquisition period. After acquiring the value of each performance parameter, the relationship between the value and its corresponding value range can be determined. If the value of the performance parameter is within the value range, it can be directly used in subsequent steps. If the value of the performance parameter is not within the value range, a target value determined based on the value range can be used as the value of the performance parameter. The value range is determined based on the acquired performance parameter values. For every N performance parameter values ​​acquired, the value range can be determined based on some or all of the acquired N performance parameter values, a preset upward floating range, and a preset downward floating range.

[0063] In the process of determining the above value range, N can be a positive integer. For example, if 10 performance parameters have been obtained, the value range can be determined based on the last 5 performance parameters, or based on the last 3 performance parameters, or based on all 10 performance parameters. Specifically, if the value range is determined by the last 5 performance parameters, it can be determined by the average value of the last 5 performance parameters out of the 10 performance parameters, and then a preset upward floating range and a preset downward floating range are added above and below the average value, respectively; or it can be determined by the median of the last 5 performance parameters out of the 10 performance parameters, and then a preset upward floating range and a preset downward floating range are added above and below the median value, respectively.

[0064] In another possible embodiment, after all the data is acquired, the preprocessing of the performance parameters can be completed in the following ways: resampling all the data, filling in missing values, using the 3-σ algorithm to detect outliers and replacing them with the average of the four points adjacent to the outlier, and using a Butterworth low-pass filter to filter the performance parameters to highlight the data trend and reduce the impact of noise.

[0065] In one possible embodiment, the type of variation pattern of multiple values ​​can be determined based on the generation time and magnitude relationship of the acquired values ​​of the same performance parameter. The variation pattern type can include periodic, stable, and trend-based. Periodic data might correspond to values ​​of 88, 90, 92, 88, 90, 92 over a period of time, with each period consisting of three values. Stable data might correspond to values ​​of 90, 90, 90, 90, 90 over a period of time, indicating stable data. Trend-based data might correspond to values ​​of 80, 81, 82, 83, 84, 85 over a period of time. It should be noted that in actual data acquisition, the values ​​may not be as accurate as in the example, but within acceptable limits, the variation pattern type of multiple values ​​can be determined based on the generation time and magnitude relationship of the acquired values ​​of the same performance parameter.

[0066] In another possible embodiment, the STL decomposition method can be used to decompose multiple values ​​of the same performance parameter into three components: periodic component, stable component, and trend component. Then, the Spearman rank correlation coefficient is used to calculate the correlation between the three components and the original data, and the same performance parameter is classified into the component type with the highest correlation coefficient. For example, if the periodic component of a certain performance parameter is 80%, the stable component is 90%, and the trend component is 99%, then the performance parameter can be considered to be a trend type performance parameter.

[0067] In one possible embodiment, before selecting the target base model pool corresponding to a specific change pattern type from multiple base model pools, the base models can be classified to determine multiple base model pools. First, multiple base models can be prepared, such as ARIMA, SARIMAX, Prophet, XGBoost, MLPLSTM, KNN, GBDT, linear regression, and random forest, etc. The specific number of base models is not limited in this application. Then, multiple sets of training data can be classified according to each change pattern type, that is, three sets of training data can be obtained: periodic training data, stable training data, and trend training data. It should be noted that each change pattern type corresponds to at least one set of training data, and each set of training data should include data to be processed and validation data. For any one of the multiple base models, at least one set of training data corresponding to each change pattern type can be input, and the fitting result corresponding to each change pattern type can be determined based on the output training results and validation data. Based on multiple fitting results, the change pattern type corresponding to any one base model can be determined. Finally, at least one base model corresponding to each change pattern type can be saved to the base model corresponding to the change pattern type to obtain multiple base model pools.

[0068] For example, for the SARIMAX base model, the SARIMAX base model is trained with periodic training data, stable training data, and trend training data respectively. The fitting result of the SARIMAX base model for periodic training data is 95%, the fitting result for stable training data is 80%, and the fitting result for trend training data is 60%. Therefore, it can be determined that the change pattern corresponding to the SARIMAX base model should be periodic. Thus, the SARIMAX base model can be saved to the base model pool corresponding to periodic data.

[0069] In one possible embodiment, after determining multiple base model pools, a target base model pool corresponding to a specific type of change pattern can be selected from the multiple base model pools.

[0070] Step S202: Input the first part of the multiple values ​​into each target base model in the target base model pool, and determine the prediction accuracy of each target base model based on the second part of the multiple values ​​and the values ​​output by each target base model.

[0071] In one possible embodiment, after obtaining the values ​​of multiple performance parameters, the obtained values ​​can be divided into two parts. The first part of the values ​​is used to process and predict for each target base model in the target base model pool. Each target base model in the target base model pool makes predictions based on the values ​​in the first part, and each target base model outputs a value. The value output by each target base model is compared with the values ​​in the second part to determine the prediction accuracy of each target base model. Typically, the number of values ​​in the first part is greater than 1, and the number of values ​​in the second part can be 1 or any number greater than 1. For example, if the obtained values ​​of multiple performance parameters are 90, 91, 92, 93, 94, 95, and 96 (a total of 7 values), and the first part of the values ​​includes the first 6 values, and 5 target base models make predictions based on the first part of the values ​​90, 91, 92, 93, 94, and 95, the target base model A outputs a value of 90, the target base model B outputs a value of 91, the target base model C outputs a value of 93, the target base model D outputs a value of 96, and the target base model E outputs a value of 95. If the first part of the value is 99, then the prediction accuracy of each target base model is determined based on the second part of the value 96. If the first part of the value contains the first 5 values, the 5 target base models make predictions based on the first part of the values ​​90, 91, 92, 93, and 94. The output values ​​of target base model A are 90 and 95, the output values ​​of target base model B are 91 and 93, the output values ​​of target base model C are 93 and 94, the output values ​​of target base model D are 96 and 97, and the output values ​​of target base model E are 99 and 101. Similarly, the output of each target base model is compared with the second part of the value to determine the prediction accuracy of each target base model.

[0072] Step S203: Select a set number of target base models based on the prediction accuracy of each target base model.

[0073] In one possible embodiment, the target base models in the target base model pool can be sorted according to their prediction accuracy for the acquired performance parameters, and a set number of target base models can be selected. For example, the prediction accuracy of each target base model for the acquired performance parameters is 70% for model A, 80% for model B, 90% for model C, 99% for model D, and 88% for model E. If the set number is 3, that is, 3 target base models are selected, models C, D, and E can be selected for subsequent steps.

[0074] In another possible embodiment, the dynamic ranking of each base model in the target base model pool can be achieved by evaluating the Symmetric Mean Absolute Percentage Error (SMAPE) of each model, which also enables the selection of a set number of target base models from the target base model pool.

[0075] In another possible embodiment, training data with different levels of fit to the type of variation pattern can be prepared in advance. For example, periodic training data A has a 60% fit to the periodic pattern, periodic training data B has a 70% fit, periodic training data C has an 80% fit, and periodic training data D has a 90% fit. Through multiple sets of training data with different fit levels, the data with the best fit among the target base models in the target base model pool corresponding to the periodic pattern can be determined. For example, target base model A is best suited to handle periodic patterns. The target base model B is best suited for data with a 90% fit to periodic patterns, while the target base model C is best suited for data with a 60% fit to periodic patterns, and the target base model D is best suited for data with a 70% fit to periodic patterns. Subsequently, the fit between the acquired performance parameters and each variation pattern type can be determined based on the multiple values ​​of the acquired performance parameters. For example, if the multiple values ​​of the acquired performance parameters have a 95% fit to periodic data, and only one target base model needs to be selected, then only target base model A can be selected to implement the subsequent steps.

[0076] Step S204: Input the second part of the numerical values ​​into a set number of target base models to obtain at least one numerical prediction result, and determine the fault prediction result based on at least one numerical prediction result.

[0077] In one possible embodiment, after selecting a set number of target base models, a second part of the numerical values ​​that are not input into the target base models can be input into the target base models to predict subsequent numerical values. After receiving the second part of the numerical values, the target base model can obtain at least one numerical prediction result. If one set number of target base models is selected, one numerical prediction result is obtained. If multiple set number of target base models are selected, multiple numerical prediction results are obtained. Each target base model can generate one numerical prediction result.

[0078] In one possible embodiment, if there is only one numerical prediction result, the fault prediction result can be determined based on the change pattern type corresponding to the single numerical prediction result; if there are multiple numerical prediction results, the multiple prediction results can be linearly fitted to obtain the final numerical prediction result, and the fault prediction result can be determined based on the final numerical prediction result.

[0079] In another possible embodiment, for each of the selected target base models, K-fold cross-validation can be performed using the second part of the data to generate multiple prediction results. The average of these multiple prediction results is then used as the final numerical prediction result. In other words, multiple numerical prediction results are obtained through multiple target base models. These multiple numerical prediction results can then be input into the meta-model XGBoost to determine the final fault prediction result. This method is particularly suitable for handling situations with high model diversity and large datasets, effectively improving the model's generalization ability and prediction accuracy.

[0080] In one possible embodiment, the fault prediction result can be determined as follows: if the change pattern of multiple values ​​is of type 1, and the difference between the final predicted value and the maximum value among multiple performance parameters is greater than a first threshold, then the fault prediction result can be determined as a fault occurring. The first type of value change pattern can include periodic or stable data. If the change pattern of multiple values ​​is of type 2, and the final predicted value is greater than a second threshold, then the fault prediction result can be determined as a fault occurring. The second type of value change pattern can include stable or trending data. If the change pattern of multiple values ​​is of type 3, and the final predicted value is not within the target confidence interval, then the fault prediction result can be determined as a fault occurring. The target confidence interval is determined based on the maximum and minimum values ​​among multiple values ​​of the same performance parameter. The third type of value change pattern can include periodic or trending data. It should be noted that different fault prediction methods can be determined based on different value change pattern types, and different fault prediction methods can be determined based on different performance parameters; this application does not impose any limitations on this.

[0081] In one possible embodiment, if the fault prediction result indicates that a fault has occurred, the user can be prompted with relevant fault information. This information may include at least one of the following: the performance parameters corresponding to the fault, the type of change pattern of the performance parameters corresponding to the fault, the value of the performance parameters corresponding to the fault, and fault remediation measures. The fault remediation measures may involve embedding an intelligent question-answering model within the computing device. The relevant fault information, such as the performance parameters corresponding to the fault, the type of change pattern of the performance parameters corresponding to the fault, and the value of the performance parameters corresponding to the fault, is input into the intelligent question-answering model. Then, the output of the intelligent question-answering model in response to the input is displayed to the user to resolve the fault.

[0082] The fault prediction method, apparatus, computing device, and storage medium provided in this application embodiment can filter target base model pools according to the numerical change patterns corresponding to different performance parameters, dynamically sort the base models in the target base model pools, select a set number of target base models with the highest prediction accuracy, use the target base models to perform numerical prediction to obtain numerical prediction results, and then determine the corresponding fault prediction results. This method can avoid the impact of changes in data distribution on the fault prediction process. The base models in different base model pools have different processing capabilities for different data change patterns. For different data distributions, that is, different data change pattern types, a base model with stronger processing capabilities can be selected to process the corresponding data change pattern type, which can ensure the accuracy of prediction, avoid fault occurrence and data loss, and improve the user experience.

[0083] In one specific embodiment, Figure 3 A detailed flowchart of a fault prediction method provided in an embodiment of this application is shown, such as... Figure 3 As shown, the method may include the following steps:

[0084] Step S301: Periodically obtain the values ​​of performance parameters.

[0085] Step S302: For each performance parameter value obtained, determine the relationship between the performance parameter value and the value range.

[0086] Step S303: If the value of the performance parameter is not within the range, then the target value determined based on the range is used as the value of the performance parameter.

[0087] In one possible embodiment, the value range is determined based on the values ​​of the acquired performance parameters in the following manner: for every N values ​​of performance parameters acquired, the value range is determined based on some or all of the acquired values ​​of the N performance parameters, a preset upward floating range, and a preset downward floating range; where N is a positive integer.

[0088] Step S304: Classify the multiple sets of training data according to each type of change pattern, so that each type of change pattern corresponds to at least one set of training data.

[0089] Step S305: For any base model, input at least one set of training data corresponding to each type of change pattern into any base model, and determine the fitting result corresponding to each type of change pattern based on the output training results and validation data.

[0090] Step S306: Determine the type of change pattern corresponding to any base model based on multiple fitting results.

[0091] Step S307: Save at least one base model corresponding to each change pattern type to the base model pool corresponding to the change pattern type.

[0092] Step S308: Based on the generation time of multiple values ​​of the same performance parameter and the size relationship of the multiple values, determine the change pattern type of the multiple values, and select the target base model pool corresponding to the determined change pattern type from multiple base model pools.

[0093] Step S309: Input the first part of the multiple values ​​into each target base model in the target base model pool, and determine the prediction accuracy of each target base model based on the second part of the multiple values ​​and the values ​​output by each target base model.

[0094] Step S310: Select a set number of target base models based on the prediction accuracy of each target base model.

[0095] Step S311: Input the second part of the numerical values ​​into a set number of target base models to obtain at least one numerical prediction result.

[0096] Step S312: Determine whether the number of digital prediction results is one. If yes, proceed to step S313; otherwise, proceed to step S314.

[0097] Step S313: Determine the fault prediction result based on the final numerical prediction result and the change pattern type corresponding to the numerical prediction result.

[0098] In one possible embodiment, the final numerical prediction result is the numerical prediction result.

[0099] Step S314: Perform linear fitting on multiple prediction results to obtain the final numerical prediction result, and determine the fault prediction result based on the final numerical prediction result and the corresponding change pattern type.

[0100] In one possible embodiment, if the variation pattern of multiple values ​​is of the first type, and the difference between the final value prediction result and the maximum value among multiple performance parameters is greater than a first threshold, then the fault prediction result is determined to be a fault occurrence.

[0101] If the change pattern of multiple values ​​is of type two, and the final value prediction result is greater than the second threshold, then the fault prediction result is determined to be a fault occurrence.

[0102] If the variation pattern of multiple values ​​is of type 3, and the final numerical prediction result is not within the target confidence interval, then the fault prediction result is determined to be a fault occurrence; the target confidence interval is determined based on the maximum and minimum values ​​among multiple values ​​of the same performance parameter.

[0103] Step S315: If the fault prediction result is determined to be a fault, then the user is prompted with fault information; the fault information includes at least one of the following: the performance parameter corresponding to the fault, the type of change pattern of the performance parameter corresponding to the fault, the value of the performance parameter corresponding to the fault, and the fault repair measures.

[0104] Based on the same inventive concept Figure 4 A structural block diagram of a fault prediction device provided in an embodiment of this application is shown below. Figure 4 As shown, the fault prediction device 400 may include:

[0105] The target base model pool determination unit 401 is used to determine the change pattern type of the multiple values ​​based on the generation time of multiple values ​​of the same performance parameter and the size relationship of the multiple values, and select the target base model pool corresponding to the determined change pattern type from the multiple base model pools.

[0106] The prediction accuracy determination unit 402 is used to input the first part of the plurality of values ​​into each target base model in the target base model pool, and determine the prediction accuracy of each target base model based on the second part of the plurality of values ​​and the values ​​output by each target base model; wherein the generation time of the first part of the values ​​is earlier than the generation time of the second part of the values.

[0107] The target base model determination unit 403 is used to select a set number of target base models based on the prediction accuracy of each target base model;

[0108] The fault prediction result determination unit 404 is used to input the second part of the numerical values ​​into the set number of target base models respectively to obtain at least one numerical prediction result, and determine the fault prediction result based on the at least one numerical prediction result.

[0109] In one possible implementation, the target base model pool determination unit 401 can also be used to periodically acquire the values ​​of performance parameters.

[0110] For each performance parameter value obtained, the relationship between the performance parameter value and the value range is determined; the value range is determined based on the obtained performance parameter values.

[0111] If the value of the performance parameter is not within the range, the target value determined based on the range will be used as the value of the performance parameter.

[0112] In one possible implementation, the target base model pool determination unit 401 can also be used to determine the value range based on some or all of the values ​​of the N performance parameters, a preset upward floating range, and a preset downward floating range for each N performance parameters obtained; wherein, N is a positive integer.

[0113] In one possible implementation, the target base model pool determination unit 401 can also be used to classify the multiple sets of training data according to each change pattern type, so that each change pattern type corresponds to at least one set of training data; each set of training data includes data to be processed and validation data;

[0114] For any base model, at least one set of training data corresponding to each of the aforementioned change pattern types is input into the base model, and the fitting result corresponding to each of the aforementioned change pattern types is determined based on the output training results and the validation data.

[0115] The type of change pattern corresponding to any base model is determined based on multiple fitting results;

[0116] At least one base model corresponding to each of the aforementioned change pattern types is saved to the base model pool corresponding to the aforementioned change pattern type.

[0117] In one possible implementation, the fault prediction result determination unit 404 is specifically used to determine the fault prediction result based on the change pattern type corresponding to the final numerical prediction result if the number of numerical prediction results obtained is one, wherein the final numerical prediction result is the numerical prediction result.

[0118] If the number of numerical prediction results obtained is greater than one, then the multiple prediction results are linearly fitted to obtain the final numerical prediction result, and the fault prediction result is determined based on the final numerical prediction result and the corresponding change pattern type.

[0119] In one possible implementation, the fault prediction result determination unit 404 is specifically used to determine that the fault prediction result is a fault has occurred if the change pattern of the plurality of values ​​is of the first type and the difference between the final value prediction result and the maximum value among the plurality of performance parameters is greater than a first threshold.

[0120] If the pattern of change of the multiple values ​​is of the second type, and the final value prediction result is greater than the second threshold, then the fault prediction result is determined to be a fault occurrence.

[0121] If the variation pattern of the multiple values ​​is of type three, and the final value prediction result is not within the target confidence interval, then the fault prediction result is determined to be a fault occurrence; the target confidence interval is determined based on the maximum and minimum values ​​among the multiple values ​​of the same performance parameter.

[0122] In one possible implementation, the fault prediction result determination unit 404 can also be used to prompt the user with fault information if the fault prediction result is determined to be a fault; the fault information includes at least one of the following: the performance parameter corresponding to the fault, the change pattern type of the performance parameter corresponding to the fault, the value of the performance parameter corresponding to the fault, and the fault repair measures.

[0123] Based on the same inventive concept, embodiments of this application provide a computing device that can implement the fault prediction method discussed above. Please refer to... Figure 5 The computing device 500 includes a memory 501, a processor 502, and a bus 503.

[0124] The memory 501 is used to store computer programs executed by the processor 502. The memory 501 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0125] Memory 501 may be volatile memory, such as random-access memory (RAM); memory 501 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 501 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 501 may be a combination of the above-mentioned memories.

[0126] The processor 502 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 502 is used to implement the fault prediction method in the above embodiments when it calls the computer program stored in the memory 501.

[0127] This application embodiment does not limit the specific connection medium between the memory 501 and the processor 502 described above. This application embodiment... Figure 5 The memory 501 and the processor 502 are connected via a bus 503, and the bus 503 is in Figure 5 The connections between other components are shown in thick lines only and are not intended to be limiting. The 503 bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0128] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium. The computer program product includes computer program code, which, when executed on a computer, causes the computer to perform any of the fault prediction methods discussed above. Since the principle by which the above-described computer-readable storage medium solves the problem is similar to that of the fault prediction method, the implementation of the above-described computer-readable storage medium can be referred to the implementation of the method; repeated details will not be elaborated further.

[0129] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0131] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0132] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0133] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A failure prediction method characterized by, include: Based on the generation time of multiple values ​​of the same performance parameter and the magnitude relationship of the multiple values, the change pattern type of the multiple values ​​is determined, and the target base model pool corresponding to the determined change pattern type is selected from multiple base model pools. The first part of the multiple values ​​is input into each target base model in the target base model pool, and the prediction accuracy of each target base model is determined based on the second part of the multiple values ​​and the values ​​output by each target base model; wherein the generation time of the first part of the values ​​is earlier than the generation time of the second part of the values. Based on the prediction accuracy of each target base model, a set number of target base models are selected; The second part of the numerical values ​​are respectively input into the set number of target base models to obtain at least one numerical prediction result, and the fault prediction result is determined based on the at least one numerical prediction result; Each base model in the base model pool is obtained through the following methods: Multiple sets of training data are classified according to different types of change patterns, so that each type of change pattern corresponds to at least one set of training data; each set of training data includes data to be processed and validation data; For any base model, at least one set of training data corresponding to each of the aforementioned change pattern types is input into the base model, and the fitting result corresponding to each of the aforementioned change pattern types is determined based on the output training results and the validation data. The type of change pattern corresponding to any base model is determined based on multiple fitting results; At least one base model corresponding to each of the aforementioned change pattern types is saved to the base model pool corresponding to the aforementioned change pattern type.

2. The method according to claim 1, characterized in that, Before determining the type of variation pattern of the multiple values ​​based on the generation time of the acquired values ​​of the same performance parameter and the magnitude relationship of the multiple values, the method further includes: Periodically obtain the values ​​of performance parameters; For each performance parameter value obtained, the relationship between the performance parameter value and the value range is determined; the value range is determined based on the obtained performance parameter values. If the value of the performance parameter is not within the range, the target value determined based on the range will be used as the value of the performance parameter.

3. The method according to claim 2, characterized in that, The value range is determined based on the obtained performance parameter values ​​in the following manner: For each N performance parameter values ​​obtained, the value range is determined based on some or all of the obtained N performance parameter values, a preset upward floating range, and a preset downward floating range; where N is a positive integer.

4. The method according to claim 1, characterized in that, Determining the fault prediction result based on the at least one numerical prediction result includes: If the number of numerical prediction results obtained is one, the fault prediction result is determined based on the change pattern type corresponding to the final numerical prediction result and the numerical prediction result, wherein the final numerical prediction result is the numerical prediction result. If the number of numerical prediction results obtained is greater than one, then the multiple prediction results are linearly fitted to obtain the final numerical prediction result, and the fault prediction result is determined based on the final numerical prediction result and the corresponding change pattern type.

5. The method according to claim 4, characterized in that, Determining the fault prediction result based on the final numerical prediction result and the corresponding change pattern type includes: If the variation pattern of the multiple values ​​is of the first type, and the difference between the final value prediction result and the maximum value among the multiple performance parameters is greater than the first threshold, then the fault prediction result is determined to be a fault occurrence. If the pattern of change of the multiple values ​​is of the second type, and the final value prediction result is greater than the second threshold, then the fault prediction result is determined to be a fault occurrence. If the variation pattern of the multiple values ​​is of type three, and the final value prediction result is not within the target confidence interval, then the fault prediction result is determined to be a fault occurrence; the target confidence interval is determined based on the maximum and minimum values ​​among the multiple values ​​of the same performance parameter.

6. The method according to claim 1, characterized in that, After determining the fault prediction result based on the at least one numerical prediction result, the method further includes: If the fault prediction result indicates that a fault has occurred, the user is prompted with fault information. The fault information includes at least one of the following: the performance parameter corresponding to the fault, the type of change pattern of the performance parameter corresponding to the fault, the value of the performance parameter corresponding to the fault, and the fault repair measures.

7. A fault prediction device, characterized in that, include: The target base model pool determination unit is used to determine the change pattern type of the multiple values ​​based on the generation time of multiple values ​​of the same performance parameter and the size relationship of the multiple values, and select the target base model pool corresponding to the determined change pattern type from multiple base model pools. The prediction accuracy determination unit is used to input the first part of the plurality of values ​​into each target base model in the target base model pool, and determine the prediction accuracy of each target base model based on the second part of the plurality of values ​​and the values ​​output by each target base model; wherein the generation time of the first part of the values ​​is earlier than the generation time of the second part of the values. The target basis model determination unit is used to select a set number of target basis models based on the prediction accuracy of each target basis model. The fault prediction result determination unit is used to input the second part of the numerical values ​​into the set number of target base models respectively to obtain at least one numerical prediction result, and determine the fault prediction result based on the at least one numerical prediction result; Each base model in the base model pool is obtained through the following methods: Multiple sets of training data are classified according to different types of change patterns, so that each type of change pattern corresponds to at least one set of training data; each set of training data includes data to be processed and validation data; For any base model, at least one set of training data corresponding to each of the aforementioned change pattern types is input into the base model, and the fitting result corresponding to each of the aforementioned change pattern types is determined based on the output training results and the validation data. The type of change pattern corresponding to any base model is determined based on multiple fitting results; At least one base model corresponding to each of the aforementioned change pattern types is saved to the base model pool corresponding to the aforementioned change pattern type.

8. A computing device, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method of any one of claims 1-6.

Images