Fault prediction method and device, storage medium and electronic equipment

Abstract

The invention discloses a fault prediction method and device, a storage medium and electronic equipment. Relates to the field of automation control, the field of artificial intelligence and other related technical fields. Performing feature extraction on the target operation information through a feature extraction unit of the target prediction model to obtain target feature information; and predicting a fault of the artificial intelligence decision-making system based on the target feature information through a prediction unit of the target prediction model to obtain a target prediction result. Through the fault prediction method and device, the problem that the accuracy of fault prediction of the artificial intelligence decision-making system is relatively low due to the fact that the fault of the artificial intelligence decision-making system is predicted in a manual mode in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the fields of automated control, artificial intelligence and other related technical fields, and more specifically, to a fault prediction method and device, a storage medium and an electronic device. Background Art

[0002] Performance monitoring and fault warning are crucial in the management of modern servers and computing clusters, especially when large-scale data centers and cloud computing platforms are involved. System performance monitoring and fault handling have become extremely complex. With the rapid development of artificial intelligence (AI) technology, the scale and complexity of data centers have grown at an unprecedented rate, resulting in traditional manual monitoring and management methods no longer being effective. When a system encounters performance bottlenecks or failures, if they cannot be predicted and handled in a timely manner, it will not only seriously affect the user experience, but may also lead to economic losses such as data loss. At present, most monitoring systems focus more on post-analysis and processing, and lack effective mechanisms to proactively predict potential problems. In addition, existing monitoring solutions are usually unable to be deeply integrated into IT infrastructure, lack the ability to adaptively adjust monitoring strategies and warning thresholds, and require frequent manual intervention, which reduces the flexibility and response speed of the system and increases the risk of business interruption.

[0003] Currently, no effective solution has been proposed to the problem that the fault prediction of artificial intelligence decision-making systems is performed manually in related technologies, resulting in relatively low accuracy of fault prediction of artificial intelligence decision-making systems. Summary of the invention

[0004] The main purpose of this application is to provide a fault prediction method and device, a storage medium and an electronic device to solve the problem in the related art that the faults of the artificial intelligence decision-making system are predicted manually, resulting in relatively low accuracy in fault prediction of the artificial intelligence decision-making system.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a fault prediction method is provided. The method includes: obtaining the target operation information of the artificial intelligence decision-making system, wherein the target operation information includes at least: the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; extracting features from the target operation information through the feature extraction unit of the target prediction model to obtain target feature information, wherein the target prediction model is iteratively optimized based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information includes at least: the response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in the historical time period; predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0006] Furthermore, before extracting features of the target operation information through a feature extraction unit of the target prediction model to obtain target feature information, the method also includes: acquiring an initial prediction model and historical operation information of the artificial intelligence decision-making system; performing time series analysis on the historical operation information to obtain a target analysis result; and iteratively optimizing the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain the target prediction model.

[0007] Furthermore, a time series analysis is performed on the historical operation information to obtain a target analysis result, including: calculating the response time information, memory usage information and video memory usage information in the historical operation information to obtain a target calculation result; analyzing the operation log information in the historical operation information to obtain error operation information corresponding to the artificial intelligence decision-making system; and analyzing based on the target calculation result and the error operation information through a time series analysis model to obtain the target analysis result.

[0008] Furthermore, feature extraction is performed on the target operation information through a feature extraction unit of a target prediction model to obtain target feature information, including: feature extraction is performed on the target operation information through a convolutional layer in the feature extraction unit to obtain initial feature information; dimension reduction is performed on the initial feature information through a pooling layer in the feature extraction unit to obtain initial feature information after dimension reduction; and the initial feature information after dimension reduction is output through an output layer in the feature extraction unit to obtain the target feature information.

[0009] Furthermore, after predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, the method also includes: determining a first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in a target database based on the target prediction result; optimizing the artificial intelligence decision-making system based on the first optimization strategy to obtain a target optimization result.

[0010] Furthermore, after optimizing the artificial intelligence decision-making system according to the first optimization strategy and obtaining the target optimization result, the method also includes: feeding back the target optimization result to the target object, and detecting whether a target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; if the target instruction is received, obtaining the target optimization information input by the target object; determining a second optimization strategy based on the target optimization information; optimizing the artificial intelligence decision-making system again based on the second optimization strategy, and storing the second optimization strategy in the target database.

[0011] Furthermore, obtaining target operating information of the artificial intelligence decision-making system includes: obtaining initial operating information of the artificial intelligence decision-making system; performing data reduction processing on the initial operating information to obtain reduced initial operating information; performing data integration processing on the reduced initial operating information to obtain integrated initial operating information; performing cleaning processing and missing value processing on the integrated initial operating information to obtain the target operating information.

[0012] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a fault prediction device is provided. The device includes: a first acquisition unit, which is used to acquire the target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; an extraction unit, which is used to extract features of the target operation information through the feature extraction unit of the target prediction model to obtain target feature information, wherein the target prediction model is obtained by iterative optimization based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: the response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in the historical time period; a prediction unit, which is used to predict the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0013] Furthermore, the device also includes: a second acquisition unit, used to obtain the initial prediction model and the historical operation information of the artificial intelligence decision-making system before extracting features of the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information; an analysis unit, used to perform time series analysis on the historical operation information to obtain a target analysis result; and a first optimization unit, used to iteratively optimize the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain the target prediction model.

[0014] Furthermore, the analysis unit includes: a calculation module, which is used to calculate the response time information, memory usage information and video memory usage information in the historical operation information to obtain the target calculation result; a first analysis module, which is used to analyze according to the operation log information in the historical operation information to obtain the error operation information corresponding to the artificial intelligence decision-making system; and a second analysis module, which is used to analyze based on the target calculation result and the error operation information through a time series analysis model to obtain the target analysis result.

[0015] Furthermore, the extraction unit includes: an extraction module, used to perform feature extraction on the target operation information through the convolution layer in the feature extraction unit to obtain initial feature information; a dimensionality reduction module, used to perform dimensionality reduction processing on the initial feature information through the pooling layer in the feature extraction unit to obtain the initial feature information after dimensionality reduction; and an output module, used to output the initial feature information after dimensionality reduction through the output layer in the feature extraction unit to obtain the target feature information.

[0016] Furthermore, the device also includes: a first determination unit, which is used to predict the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model, and after obtaining the target prediction result, determine the first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in the target database according to the target prediction result; a second optimization unit, which is used to optimize the artificial intelligence decision-making system according to the first optimization strategy to obtain the target optimization result.

[0017] Furthermore, the device also includes: a feedback unit, which is used to optimize the artificial intelligence decision-making system according to the first optimization strategy and obtain the target optimization result, then feed back the target optimization result to the target object, and detect whether a target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; a third acquisition unit, which is used to obtain the target optimization information input by the target object if the target instruction is received; a second determination unit, which is used to determine a second optimization strategy based on the target optimization information; a third optimization unit, which is used to optimize the artificial intelligence decision-making system again based on the second optimization strategy, and store the second optimization strategy in the target database.

[0018] Furthermore, the first acquisition unit includes: an acquisition module, used to acquire the initial operation information of the artificial intelligence decision-making system; a first processing module, used to perform data reduction processing on the initial operation information to obtain the reduced initial operation information; a second processing module, used to perform data integration processing on the reduced initial operation information to obtain the integrated initial operation information; a third processing module, used to perform cleaning processing and missing value processing on the integrated initial operation information to obtain the target operation information.

[0019] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer-readable storage medium is provided, wherein the storage medium stores a program, wherein when the program is running, the device where the storage medium is located is controlled to execute any one of the above-mentioned fault prediction methods.

[0020] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic device is further provided, the electronic device includes one or more processors and a memory, and the memory is used to store one or more processors to implement any one of the above-mentioned fault prediction methods.

[0021] Through this application, the following steps are adopted: obtaining the target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; extracting the features of the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information, wherein the target prediction model is obtained by iterative optimization based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: the response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in the historical time period; predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain the target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system, and through this application, the problem of low accuracy of fault prediction of the artificial intelligence decision-making system caused by artificial means in the related technology is solved. In this scheme, the historical operation information of the artificial intelligence decision-making system is studied and analyzed to obtain the target prediction model, and the feature extraction unit of the target prediction model is used to extract the features of the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system to obtain the target feature information. According to the target feature information, the prediction unit of the target prediction model predicts the failure of the artificial intelligence decision-making system to obtain the target prediction result. By continuously tracking the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system, and using AI technology to analyze it in real time, predicting and identifying the early signs of failure and performance degradation trends of the system, the stability of the system is improved, and the dependence on professional maintenance personnel is reduced, thereby achieving the effect of improving the accuracy of system failure prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 is a flowchart of a fault prediction method provided according to an embodiment of the present application;

[0024] Figure 2 is a schematic diagram of a fault prediction method provided according to an embodiment of the present application;

[0025] Figure 3 is a schematic diagram of a fault prediction device provided according to an embodiment of the present application;

[0026] Figure 4 is a schematic diagram of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set between this system and the relevant user or organization. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0031] The present invention is described below in conjunction with preferred implementation steps. Figure 1 is a flowchart of a fault prediction method provided according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0032] Step S101, obtaining target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system.

[0033] Optionally, the target operation data of the AI ​​system (i.e., artificial intelligence decision-making system) is collected through the data collector unit in the RPA (Robotic process automation) data procurement module, including but not limited to: response time information, prediction accuracy information, resource utilization information (for example: memory usage information and video memory usage information) and other key performance indicators of the AI ​​system.

[0034] It should be noted that the target operation data effectively ensures that the fault prediction system can quickly respond to performance fluctuations of the AI ​​system and promptly detect potential problems in the AI ​​system.

[0035] Step S102, extracting features of the target operation information through the feature extraction unit of the target prediction model to obtain target feature information, wherein the target prediction model is iteratively optimized based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in a historical time period.

[0036] Optionally, the feature extraction unit of the predictive analyzer unit (i.e., the target prediction model) in the AI ​​analysis engine module extracts key information from the target operation information and screens out the most important data features to obtain target feature information for performance analysis of the artificial intelligence decision-making system and prediction of abnormal situations.

[0037] Step S103, predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0038] Optionally, the prediction unit of the target prediction model is used to predict possible problems or performance degradation of the artificial intelligence decision-making system in the future based on target feature information and in combination with time series analysis and reinforcement learning algorithms.

[0039] In summary, the target prediction model is obtained by learning and analyzing the historical operation information of the artificial intelligence decision-making system, and the feature extraction unit of the target prediction model is used to extract the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system to obtain the target feature information. According to the target feature information, the prediction unit of the target prediction model predicts the failure of the artificial intelligence decision-making system to obtain the target prediction result. By continuously tracking the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system, and using AI technology to analyze it in real time, predicting and identifying the early signs of failure and performance degradation trends of the system, the stability of the system is improved, and the dependence on professional maintenance personnel is reduced, thereby achieving the effect of improving the accuracy of system fault prediction.

[0040] Optionally, in the fault prediction method provided in the embodiment of the present application, before performing feature extraction on the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information, the method also includes: obtaining historical operation information of the initial prediction model and the artificial intelligence decision-making system; performing time series analysis on the historical operation information to obtain a target analysis result; and iteratively optimizing the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain a target prediction model.

[0041] In an optional embodiment, a large amount of historical operation information for training and optimizing the initial prediction model is collected through a data collector unit, including but not limited to: operation error information, operation warning information, response time information, memory usage information, video memory usage information and other system operation log information of the artificial intelligence decision-making system in a historical time period. Then, a time series analysis is performed on the historical operation information to identify the failure modes and failure trends in the historical operation information, so as to predict the future behavior of the artificial intelligence decision-making system, including but not limited to: predicting when the artificial intelligence decision-making system may reach a performance bottleneck or fail, etc., to obtain the target analysis results. Finally, the initial prediction model is trained and iteratively optimized based on the target analysis results through the model trainer unit (i.e., reinforcement learning algorithm) in the AI ​​analysis engine module to determine the best target prediction model.

[0042] It should be noted that, taking the prediction of the performance of a large AI server cluster as an example: the data collector unit collects the performance indicators of the server in real time, such as CPU and memory usage, etc., and cleans and extracts features, and then passes the data to the predictive analyzer unit. In the predictive analyzer unit, the time series analysis model is used to identify abnormal patterns, such as sudden peaks in CPU usage, which may indicate an impending failure. At the same time, the reinforcement learning algorithm is used to dynamically select the most appropriate prediction model to optimize the accuracy of the prediction based on the current server load and historical failure data. For example, if historical data shows that the AI ​​server is more likely to fail under high load, the reinforcement learning algorithm may adjust its strategy to pay more attention to the prediction model under these conditions. In this way, when real-time data indicates that the AI ​​server is about to enter a high load state, the system can take measures in advance and feedback the information for subsequent processing.

[0043] By performing time series analysis on historical operating information through reinforcement learning algorithms, we can continuously learn from the actual data fed back by the system and automatically adjust the prediction strategy to accurately determine the target prediction model, thereby improving the accuracy and generalization ability of the target prediction model.

[0044] Optionally, in the fault prediction method provided in the embodiment of the present application, a time series analysis is performed on the historical operation information to obtain a target analysis result, including: calculating the response time information, memory usage information, and video memory usage information in the historical operation information to obtain a target calculation result; analyzing the operation log information in the historical operation information to obtain error operation information corresponding to the artificial intelligence decision-making system; and analyzing based on the target calculation result and the error operation information through a time series analysis model to obtain a target analysis result.

[0045] In an optional embodiment, statistical indicators such as the average, median, maximum and minimum values ​​are calculated for the response time information, memory usage information and video memory usage information in the historical operation information to obtain the target calculation results. And analysis is performed based on the operation log information in the historical operation information to obtain error operation information within the historical time period, including but not limited to: the frequency, time and related conditions of AI system error operation, so as to identify system parts or operations that may cause errors. By constructing time series models, including but not limited to: ARIMA, seasonal decomposition and other models, the trend and periodicity of the response time are analyzed to obtain the target analysis results.

[0046] By analyzing response time, memory, and video memory usage, you can identify performance bottlenecks and optimize them to improve the overall efficiency of the system. At the same time, by analyzing error logs, you can predict possible system problems and fix them before they affect users.

[0047] Optionally, in the fault prediction method provided in the embodiment of the present application, feature extraction is performed on the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information, including: feature extraction is performed on the target operation information through the convolution layer in the feature extraction unit to obtain initial feature information; dimension reduction is performed on the initial feature information through the pooling layer in the feature extraction unit to obtain initial feature information after dimension reduction; and the initial feature information after dimension reduction is output through the output layer in the feature extraction unit to obtain target feature information.

[0048] In an optional embodiment, the input target operation information is scanned in a local area through multiple filters (also called convolution kernels or feature detectors) of the convolution layer in the feature extraction unit to identify and extract features. These filters slide on the input data, calculate the weighted sum of the coverage area, and increase the nonlinear expression ability of the model through a nonlinear activation function (for example: ReLU function), so that the network can learn more complex features to obtain initial feature information. Then, the spatial dimension of the initial feature information is reduced through the pooling layer in the feature extraction unit, the number of parameters and the amount of calculation are reduced, and the abstract ability and anti-noise ability of the model are increased to obtain the initial feature information after dimensionality reduction. Finally, after a series of convolutional layers and pooling layers, the initial feature information after dimensionality reduction will be sent to the output layer, which is usually a fully connected layer followed by an activation function to generate the final target feature information.

[0049] It should be noted that common pooling operations include maximum pooling and average pooling.

[0050] Through the convolution layer, pooling layer and output layer in the feature extraction unit, the target feature information can be obtained quickly and accurately.

[0051] Optionally, in the fault prediction method provided in the embodiment of the present application, after predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain the target prediction result, the method also includes: determining a first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in the target database based on the target prediction result; optimizing the artificial intelligence decision-making system based on the first optimization strategy to obtain the target optimization result.

[0052] In an optional embodiment, after obtaining the target prediction result, the target prediction result is used as the input of the active warning and automatic response module, so that the active warning and automatic response module can make actual decisions based on this information and guide the system on how to respond. The active warning and automatic response module includes: a decision manager unit, an execution coordinator unit and a repair executor unit. The decision manager unit is responsible for formulating a solution strategy (i.e., the first optimization strategy) according to the target prediction result, the execution coordinator unit is responsible for specific implementation, and the repair executor unit is responsible for implementing specific repair actions.

[0053] It should be noted that the role of the decision manager unit is to make decisions based on the target prediction results and other relevant information, and to formulate corresponding action plans. This unit usually contains logic and rules to evaluate the severity of the target prediction results and determine the most appropriate response strategy. For example: If it is predicted that the server may have performance problems, the decision manager unit needs to decide whether to restart the service, reallocate resources, or perform other maintenance tasks. This unit will consider a variety of factors, such as: current system load, historical maintenance records, resource availability, etc., to determine the best course of action. Take the AI ​​system server running a critical data analysis task as an example: if the target prediction result predicts that the server's response time will increase significantly in the next few hours due to high load by analyzing the server's performance data. This prediction result is passed to the decision manager unit, which, after evaluating the current system status and resource allocation, decides to automatically expand server resources (for example: start additional computing instances) to share the load.

[0054] It should be noted that the role of the execution coordinator unit is to ensure that the response measures formulated by the decision manager unit are effectively executed and to optimize the task scheduling algorithm to reduce system downtime and improve operating efficiency. The execution coordinator unit can implement a variety of functions, including but not limited to: task scheduling optimization, resource allocation, execution monitoring and exception handling and feedback mechanism. In the AI ​​server, task scheduling optimization is achieved through an intelligent scheduling architecture, which manages and optimizes resource allocation based on artificial intelligence algorithms and data analysis technology. For example, by real-time monitoring of task requirements and resource status, the intelligent scheduling system can automatically adjust the task allocation plan to improve resource utilization and work efficiency. In the AI ​​server environment, the execution coordinator unit needs to ensure that the resources required to execute the task are reasonably allocated. This may involve the dynamic allocation of GPU, CPU and other hardware resources to ensure that the task can run efficiently. The execution coordinator unit monitors the execution status of the task to ensure that each task can be carried out according to the predetermined plan. During the execution of the task, if an abnormal situation is encountered, the execution coordinator unit can make timely adjustments or reschedules to ensure the continuity of the task and the stability of the system. The execution coordinator unit feeds back the results of the task execution to the decision manager unit for subsequent decision adjustments. This feedback mechanism is critical to the continuous optimization and adaptive capabilities of the system.

[0055] It should be noted that, for example, an AI server cluster is running multiple machine learning training tasks: each task has different requirements for computing resources. The execution coordinator unit dynamically allocates GPUs and other computing resources based on the priority and resource requirements of each task. If a task runs slowly due to insufficient resources, the execution coordinator unit may adjust the resource allocation of other low-priority tasks to ensure that critical tasks have sufficient computing power. In addition, if a GPU in the server fails, the execution coordinator unit will immediately remove the GPU from the resource pool to prevent tasks from being assigned to the faulty hardware, and reschedule the affected tasks to other healthy GPUs to minimize the impact on the training tasks.

[0056] It should be noted that the repair actions of the repair executor unit include but are not limited to: restarting services, reconfiguring system parameters, updating software or hardware drivers, etc. The repair executor unit specifically performs the corresponding repair steps according to the tasks assigned by the task scheduler unit. For example: Suppose a node in an AI system server cluster has degraded hardware performance. The decision manager unit analyzes the fault and decides that the service of the node needs to be restarted. After receiving this decision, the execution coordinator unit will schedule the restart task and ensure that it is executed within the appropriate time window to reduce the impact on user services. The repair executor unit will then perform the restart operation, and after the operation is completed, it will feed back the results to the execution coordinator unit, which will then report the results to the decision manager unit to confirm whether the problem has been resolved.

[0057] Through the optimization and coordination of the decision manager unit, the execution coordinator unit, and the repair executor unit, efficient fault handling of the AI ​​system is achieved, enabling it to maintain high efficiency and high availability, and maintain stable operation even in the face of hardware failures or resource competition.

[0058] Optionally, in the fault prediction method provided in the embodiment of the present application, after optimizing the artificial intelligence decision-making system according to the first optimization strategy to obtain the target optimization result, the method also includes: feeding back the target optimization result to the target object, and detecting whether the target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; if the target instruction is received, obtaining the target optimization information input by the target object; determining the second optimization strategy based on the target optimization information; optimizing the artificial intelligence decision-making system again based on the second optimization strategy, and storing the second optimization strategy in the target database.

[0059] In an optional embodiment, after the repair task of the result feedback mechanism unit in the automated repair execution module is completed, the execution results (i.e., target optimization results) are collected and fed back. These results include, but are not limited to: whether the repair is successful, whether the system performance has returned to normal levels, and issues that may still need further processing. These results can also be fed back to the user (i.e., the target object) through the information reminder and interaction module.

[0060] It should be noted that the information reminder and interaction module provides users with an intuitive operation interface and ensures the efficiency of the information notification and feedback process. Most of the time, users do not need to use this module, because many abnormalities can be repaired by the system through the automatic repair execution module. Users mainly use the capabilities of this module to check and understand afterwards; at the same time, for some problems that the automatic repair execution module cannot handle independently, this module can promptly remind relevant staff.

[0061] It should be noted that the dashboard display unit in the information reminder and interaction module provides a window for users to interact with the system, providing real-time data visualization services and system status overviews. For example: it can display the current system response time, prediction accuracy, resource utilization, and whether the user has triggered the target instruction that has not adopted the execution result, so that users can quickly understand the system operation status and make corresponding decisions. After monitoring that the user has not adopted the target instruction of the execution result, it is necessary to redefine the optimization direction and optimization steps (i.e., the second optimization strategy) of the AI ​​system based on the optimization opinions (i.e., the target optimization information) input by the user, and re-optimize the AI ​​system based on the above content.

[0062] It should be noted that the alarm manager unit in the information reminder and interaction module monitors whether there are abnormal conditions in the AI ​​system optimization process. When abnormal conditions are triggered, the alarm manager unit will automatically generate an alarm. These alarms will be distributed to relevant staff to ensure that they can understand and respond to possible problems or failures in a timely manner. The methods of alarm notification can be diversified, including but not limited to: SMS, email and other communication tools to meet different notification needs.

[0063] It should be noted that detailed performance and fault reports are generated regularly or according to user needs through the report generator unit in the information reminder and interaction module. These reports provide valuable data support for relevant staff to help them analyze system performance trends, identify potential problems, and formulate corresponding optimization measures. The content of the report includes but is not limited to: all key performance indicators and event logs to ensure the comprehensiveness and accuracy of the information.

[0064] Through multiple functional units, it is possible to accurately determine whether the user is satisfied with the current target optimization results, so as to process the AI ​​system more efficiently.

[0065] Optionally, in the fault prediction method provided in the embodiment of the present application, obtaining the target operating information of the artificial intelligence decision-making system includes: obtaining the initial operating information of the artificial intelligence decision-making system; performing data reduction processing on the initial operating information to obtain reduced initial operating information; performing data integration processing on the reduced initial operating information to obtain integrated initial operating information; performing cleaning processing and missing value processing on the integrated initial operating information to obtain target operating information.

[0066] In an optional embodiment, the initial operation information of the artificial intelligence decision system is obtained through the data collector unit of the RPA data procurement module, and the data preprocessor unit of the RPA data procurement module performs dimension reduction, feature selection or data compression on it to reduce the complexity of the data, while retaining key information as much as possible to obtain the reduced initial operation information. If the reduced initial operation information comes from multiple systems or functional modules, these scattered data are integrated together through operations such as data fusion, data alignment and data merging to form a unified data set to obtain the integrated initial operation information. Finally, the data is quality checked to identify and process errors, outliers or inconsistent data, including but not limited to: removing duplicate records, correcting errors, standardizing data formats, deleting records with missing values, using mean or median filling, or using more complex statistical methods for prediction and filling, etc., to obtain the final target operation information.

[0067] It should be noted that the target operation information is stored in the data storage interface unit of the RPA data procurement module, which securely saves the collected data through the data persistence function and prevents data loss through the backup and recovery mechanism. In addition, the data storage interface unit also has index optimization and fast retrieval functions, so that data can quickly find the required information like the index system in the library. In addition, the data storage interface unit is also responsible for data integration and updating, ensuring that all systems use the latest data and managing historical versions of data. And by implementing strict access control and encryption measures, ensure data security and permission management to prevent the leakage of sensitive data.

[0068] By performing data reduction, data integration, cleaning and missing value processing on the initial operation information, the impact of abnormal data on the operation information can be quickly eliminated, and the target operation information can be accurately obtained.

[0069] It should be noted that Figure 2 Schematic diagram of a fault prediction method provided according to an embodiment of the present application. Figure 2As shown, the fault prediction method includes but is not limited to multiple functional modules: RPA data procurement module (including: data collector unit, data preprocessor unit and data storage interface unit), AI analysis engine module (including: prediction analyzer unit and model trainer unit), active warning and automatic response module (including: decision manager unit, execution coordinator unit and repair executor unit), automated repair execution module (including: result feedback mechanism unit) and information reminder and interaction module (including: dashboard display unit, alarm manager unit and report generator unit). The data collector unit is responsible for collecting real-time operation data and historical data, and providing a large amount of multimodal data for establishing pattern classification and prediction models. The data preprocessor unit performs preprocessing operations such as cleaning and filling in missing data on the received data to ensure the accuracy and integrity of the data. The data storage interface unit stores the processed data securely, retrieves and updates it quickly, and ensures that the data is not accessed by unauthorized persons. The prediction analyzer unit is responsible for predicting possible problems in the future, while the model trainer unit is responsible for training and updating machine learning models to optimize the analysis process. The core function of the active warning and automatic response module is to proactively issue warnings and execute automated processes to initiate corresponding response measures when the system detects potential performance issues or signs of failure. The automated repair execution module is responsible for automatically executing specific repair operations after the system confirms the problem.

[0070] The fault prediction method provided in the embodiment of the present application obtains the target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; the target operation information is feature extracted by the feature extraction unit of the target prediction model to obtain the target feature information, wherein the target prediction model is obtained by iterative optimization based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: the response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in the historical time period; the fault of the artificial intelligence decision-making system is predicted based on the target feature information by the prediction unit of the target prediction model to obtain the target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system, and solves the problem that the fault of the artificial intelligence decision-making system is predicted by artificial means in the related technology, resulting in a relatively low accuracy of fault prediction for the artificial intelligence decision-making system. In summary, in this scheme, the historical operation information of the artificial intelligence decision-making system is studied and analyzed to obtain the target prediction model, and the feature extraction unit of the target prediction model is used to extract the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system to obtain the target feature information. According to the target feature information, the prediction unit of the target prediction model predicts the failure of the artificial intelligence decision-making system to obtain the target prediction result. By continuously tracking the target operation information such as the response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system, and using AI technology to analyze it in real time, predicting and identifying the early signs of failure and performance degradation trends of the system, the stability of the system is improved, and the dependence on professional maintenance personnel is reduced, thereby achieving the effect of improving the accuracy of system failure prediction.

[0071] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0072] The embodiment of the present application also provides a fault prediction device. It should be noted that the fault prediction device of the embodiment of the present application can be used to execute the fault prediction method provided by the embodiment of the present application. The fault prediction device provided by the embodiment of the present application is introduced below.

[0073] Figure 3 Schematic diagram of a fault prediction device according to an embodiment of the present application. Figure 3 As shown, the device includes: a first acquisition unit 301, an extraction unit 302, and a prediction unit 303.

[0074] The first acquisition unit 301 is used to acquire target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system;

[0075] An extraction unit 302 is used to extract features from the target operation information through a feature extraction unit of a target prediction model to obtain target feature information, wherein the target prediction model is obtained by iterative optimization based on historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in a historical time period;

[0076] The prediction unit 303 is used to predict the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0077] In the fault prediction device provided in the embodiment of the present application, the first acquisition unit 301 acquires the target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; the extraction unit 302 extracts features from the target operation information through the feature extraction unit of the target prediction model to obtain target feature information, wherein the target prediction model is iteratively optimized based on the historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system within a historical time period; the prediction unit 303 predicts the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system, thereby solving the problem in the related art that the fault of the artificial intelligence decision-making system is predicted manually, resulting in a relatively low accuracy in fault prediction of the artificial intelligence decision-making system. In this solution, the historical operation information of the artificial intelligence decision-making system is studied and analyzed to obtain a target prediction model, and the feature extraction unit of the target prediction model is used to extract features of the target operation information such as the response time information, memory usage information, and video memory usage information of the artificial intelligence decision-making system to obtain the target feature information. Based on the target feature information, the prediction unit of the target prediction model predicts the failure of the artificial intelligence decision-making system to obtain the target prediction result. By continuously tracking the target operation information such as the response time information, memory usage information, and video memory usage information of the artificial intelligence decision-making system, and using AI technology to analyze it in real time, predicting and identifying the early signs of failure and performance degradation trends of the system, the stability of the system is improved, and the dependence on professional maintenance personnel is reduced, thereby achieving the effect of improving the accuracy of system failure prediction.

[0078] Optionally, in the fault prediction device provided in the embodiment of the present application, the device also includes: a second acquisition unit, used to obtain historical operation information of the initial prediction model and the artificial intelligence decision-making system before performing feature extraction on the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information; an analysis unit, used to perform time series analysis on the historical operation information to obtain a target analysis result; and a first optimization unit, used to iteratively optimize the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain a target prediction model.

[0079] Optionally, in the fault prediction device provided in the embodiment of the present application, the analysis unit includes: a calculation module, which is used to calculate the response time information, memory usage information and video memory usage information in the historical operation information to obtain the target calculation result; a first analysis module, which is used to analyze according to the operation log information in the historical operation information to obtain the error operation information corresponding to the artificial intelligence decision-making system; and a second analysis module, which is used to analyze based on the target calculation result and the error operation information through a time series analysis model to obtain the target analysis result.

[0080] Optionally, in the fault prediction device provided in the embodiment of the present application, the extraction unit includes: an extraction module, used to perform feature extraction on the target operation information through the convolution layer in the feature extraction unit to obtain initial feature information; a dimensionality reduction module, used to perform dimensionality reduction processing on the initial feature information through the pooling layer in the feature extraction unit to obtain the initial feature information after dimensionality reduction; and an output module, used to output the initial feature information after dimensionality reduction through the output layer in the feature extraction unit to obtain target feature information.

[0081] Optionally, in the fault prediction device provided in the embodiment of the present application, the device also includes: a first determination unit, used to predict the fault of the artificial intelligence decision-making system based on target feature information through the prediction unit of the target prediction model, and after obtaining the target prediction result, determine the first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in the target database according to the target prediction result; a second optimization unit, used to optimize the artificial intelligence decision-making system according to the first optimization strategy to obtain the target optimization result.

[0082] Optionally, in the fault prediction device provided in the embodiment of the present application, the device also includes: a feedback unit, which is used to optimize the artificial intelligence decision-making system according to the first optimization strategy, and after obtaining the target optimization result, feed back the target optimization result to the target object, and detect whether the target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; a third acquisition unit, which is used to obtain the target optimization information input by the target object if the target instruction is received; a second determination unit, which is used to determine the second optimization strategy based on the target optimization information; a third optimization unit, which is used to optimize the artificial intelligence decision-making system again based on the second optimization strategy, and store the second optimization strategy in the target database.

[0083] Optionally, in the fault prediction device provided in the embodiment of the present application, the first acquisition unit includes: an acquisition module, used to acquire the initial operation information of the artificial intelligence decision-making system; a first processing module, used to perform data reduction processing on the initial operation information to obtain the reduced initial operation information; a second processing module, used to perform data integration processing on the reduced initial operation information to obtain the integrated initial operation information; a third processing module, used to perform cleaning processing and missing value processing on the integrated initial operation information to obtain the target operation information.

[0084] The fault prediction device includes a processor and a memory. The first acquisition unit 301, extraction unit 302, prediction unit 303, etc. are all stored in the memory as program units. The processor executes the program units stored in the memory to achieve accurate prediction of faults.

[0085] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and accurate prediction of faults can be achieved by adjusting kernel parameters.

[0086] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0087] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a fault prediction method is implemented.

[0088] An embodiment of the present invention provides a processor, which is used to run a program, wherein a fault prediction method is executed when the program is running.

[0089] like Figure 4As shown, an embodiment of the present invention provides an electronic device, the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: obtaining target operation information of an artificial intelligence decision-making system, wherein the target operation information includes at least: response time information, memory usage information, and video memory usage information of the artificial intelligence decision-making system; performing feature extraction on the target operation information through a feature extraction unit of a target prediction model to obtain target feature information, wherein the target prediction model is iteratively optimized based on historical operation information of the artificial intelligence decision-making system, and the historical operation information includes at least: response time information, memory usage information, video memory usage information, and operation log information of the artificial intelligence decision-making system within a historical time period; predicting faults of the artificial intelligence decision-making system based on the target feature information through a prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0090] Optionally, before extracting features of the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information, the method also includes: obtaining historical operation information of the initial prediction model and the artificial intelligence decision-making system; performing time series analysis on the historical operation information to obtain a target analysis result; and iteratively optimizing the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain a target prediction model.

[0091] Optionally, performing a time series analysis on historical operation information to obtain a target analysis result includes: calculating the response time information, memory usage information and video memory usage information in the historical operation information to obtain a target calculation result; analyzing the operation log information in the historical operation information to obtain error operation information corresponding to the artificial intelligence decision-making system; and analyzing based on the target calculation result and the error operation information through a time series analysis model to obtain a target analysis result.

[0092] Optionally, feature extraction is performed on the target operation information through a feature extraction unit of the target prediction model to obtain the target feature information, including: feature extraction is performed on the target operation information through a convolutional layer in the feature extraction unit to obtain initial feature information; dimension reduction is performed on the initial feature information through a pooling layer in the feature extraction unit to obtain initial feature information after dimension reduction; and the initial feature information after dimension reduction is output through an output layer in the feature extraction unit to obtain target feature information.

[0093] Optionally, after predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain the target prediction result, the method also includes: determining a first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in the target database based on the target prediction result; optimizing the artificial intelligence decision-making system based on the first optimization strategy to obtain the target optimization result.

[0094] Optionally, after optimizing the artificial intelligence decision-making system according to the first optimization strategy and obtaining the target optimization result, the method also includes: feeding back the target optimization result to the target object, and detecting whether a target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; if the target instruction is received, obtaining the target optimization information input by the target object; determining a second optimization strategy based on the target optimization information; optimizing the artificial intelligence decision-making system again based on the second optimization strategy, and storing the second optimization strategy in the target database.

[0095] Optionally, obtaining target operating information of the artificial intelligence decision-making system includes: obtaining initial operating information of the artificial intelligence decision-making system; performing data reduction processing on the initial operating information to obtain reduced initial operating information; performing data integration processing on the reduced initial operating information to obtain integrated initial operating information; performing cleaning processing and missing value processing on the integrated initial operating information to obtain target operating information.

[0096] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0097] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: obtaining target operation information of an artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information, and video memory usage information of the artificial intelligence decision-making system; performing feature extraction on the target operation information through a feature extraction unit of a target prediction model to obtain target feature information, wherein the target prediction model is iteratively optimized based on historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information, and operation log information of the artificial intelligence decision-making system within a historical time period; predicting faults of the artificial intelligence decision-making system based on the target feature information through a prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

[0098] Optionally, before extracting features of the target operation information through the feature extraction unit of the target prediction model to obtain the target feature information, the method also includes: obtaining historical operation information of the initial prediction model and the artificial intelligence decision-making system; performing time series analysis on the historical operation information to obtain a target analysis result; and iteratively optimizing the initial prediction model based on the target analysis result through a reinforcement learning algorithm to obtain a target prediction model.

[0099] Optionally, performing a time series analysis on historical operation information to obtain a target analysis result includes: calculating the response time information, memory usage information and video memory usage information in the historical operation information to obtain a target calculation result; analyzing the operation log information in the historical operation information to obtain error operation information corresponding to the artificial intelligence decision-making system; and analyzing based on the target calculation result and the error operation information through a time series analysis model to obtain a target analysis result.

[0100] Optionally, feature extraction is performed on the target operation information through a feature extraction unit of the target prediction model to obtain the target feature information, including: feature extraction is performed on the target operation information through a convolutional layer in the feature extraction unit to obtain initial feature information; dimension reduction is performed on the initial feature information through a pooling layer in the feature extraction unit to obtain initial feature information after dimension reduction; and the initial feature information after dimension reduction is output through an output layer in the feature extraction unit to obtain target feature information.

[0101] Optionally, after predicting the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain the target prediction result, the method also includes: determining a first optimization strategy corresponding to the artificial intelligence decision-making system from multiple optimization strategies in the target database based on the target prediction result; optimizing the artificial intelligence decision-making system based on the first optimization strategy to obtain the target optimization result.

[0102] Optionally, after optimizing the artificial intelligence decision-making system according to the first optimization strategy and obtaining the target optimization result, the method also includes: feeding back the target optimization result to the target object, and detecting whether a target instruction is received, wherein the target instruction is used to characterize that the target object has not adopted the target optimization result; if the target instruction is received, obtaining the target optimization information input by the target object; determining a second optimization strategy based on the target optimization information; optimizing the artificial intelligence decision-making system again based on the second optimization strategy, and storing the second optimization strategy in the target database.

[0103] Optionally, obtaining target operating information of the artificial intelligence decision-making system includes: obtaining initial operating information of the artificial intelligence decision-making system; performing data reduction processing on the initial operating information to obtain reduced initial operating information; performing data integration processing on the reduced initial operating information to obtain integrated initial operating information; performing cleaning processing and missing value processing on the integrated initial operating information to obtain target operating information.

[0104] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0105] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0106] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0108] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0109] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0110] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0111] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0112] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0113] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A fault prediction method, characterized in that: include: Obtain target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; The target operation information is subjected to feature extraction by a feature extraction unit of a target prediction model to obtain target feature information, wherein the target prediction model is obtained by iterative optimization based on historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in a historical time period; The prediction unit of the target prediction model predicts the fault of the artificial intelligence decision-making system based on the target feature information to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

2. The method according to claim 1, characterized in that Before extracting features from the target operation information by a feature extraction unit of the target prediction model to obtain target feature information, the method further includes: Obtaining the initial prediction model and historical operation information of the artificial intelligence decision-making system; Performing time series analysis on the historical operation information to obtain target analysis results; The initial prediction model is iteratively optimized based on the target analysis result through a reinforcement learning algorithm to obtain the target prediction model.

3. The method according to claim 2, characterized in that Performing time series analysis on the historical operation information to obtain target analysis results includes: Calculate the response time information, memory usage information, and video memory usage information in the historical operation information to obtain a target calculation result; Analyze the operation log information in the historical operation information to obtain error operation information corresponding to the artificial intelligence decision-making system; The target analysis result is obtained by analyzing the target calculation result and the erroneous operation information through a time series analysis model.

4. The method according to claim 1, characterized in that The target operation information is subjected to feature extraction by a feature extraction unit of a target prediction model, and the target feature information obtained includes: Extracting features of the target operation information through a convolutional layer in the feature extraction unit to obtain initial feature information; Performing dimensionality reduction processing on the initial feature information through a pooling layer in the feature extraction unit to obtain initial feature information after dimensionality reduction; The initial feature information after dimensionality reduction is outputted through the output layer in the feature extraction unit to obtain the target feature information.

5. The method according to claim 1, characterized in that After predicting the fault of the artificial intelligence decision-making system based on the target feature information by the prediction unit of the target prediction model to obtain the target prediction result, the method further includes: Determining a first optimization strategy corresponding to the artificial intelligence decision-making system from a plurality of optimization strategies in a target database according to the target prediction result; The artificial intelligence decision-making system is optimized according to the first optimization strategy to obtain a target optimization result.

6. The method according to claim 5, characterized in that After optimizing the artificial intelligence decision-making system according to the first optimization strategy to obtain a target optimization result, the method further includes: Feeding back the target optimization result to the target object, and detecting whether a target instruction is received, wherein the target instruction is used to indicate that the target object has not adopted the target optimization result; If the target instruction is received, obtaining the target optimization information input by the target object; Determining a second optimization strategy according to the target optimization information; According to the second optimization strategy, the artificial intelligence decision-making system is optimized again, and the second optimization strategy is stored in the target database.

7. The method according to claim 1, characterized in that Obtaining the target operation information of the artificial intelligence decision-making system includes: Obtaining initial operation information of the artificial intelligence decision-making system; Performing data reduction processing on the initial operation information to obtain reduced initial operation information; Performing data integration processing on the reduced initial operation information to obtain integrated initial operation information; The integrated initial operation information is cleaned and missing values ​​are processed to obtain the target operation information.

8. A fault prediction device, characterized in that: include: A first acquisition unit is used to acquire target operation information of the artificial intelligence decision-making system, wherein the target operation information at least includes: response time information, memory usage information and video memory usage information of the artificial intelligence decision-making system; An extraction unit is used to extract features from the target operation information through a feature extraction unit of a target prediction model to obtain target feature information, wherein the target prediction model is obtained by iterative optimization based on historical operation information of the artificial intelligence decision-making system, and the historical operation information at least includes: response time information, memory usage information, video memory usage information and operation log information of the artificial intelligence decision-making system in a historical time period; A prediction unit is used to predict the fault of the artificial intelligence decision-making system based on the target feature information through the prediction unit of the target prediction model to obtain a target prediction result, wherein the target prediction result is used to optimize the artificial intelligence decision-making system.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the storage medium is controlled to execute the fault prediction method according to any one of claims 1 to 7 on a device.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the fault prediction method described in any one of claims 1 to 7.

Images