Data evaluation model generation method and device, abnormal data monitoring method and device

Abstract

The disclosed embodiment discloses a data evaluation model generation method and device, an abnormal data monitoring method and device, a computer-readable storage medium and an electronic device, wherein the abnormal data monitoring method comprises: obtaining a data set to be processed from a data storage area, wherein the data storage area is used to store data generated by the target software during operation; performing serialization preprocessing on the data set to be processed to obtain a serialized data set; and processing the serialized data set based on a pre-trained data evaluation model to obtain description information for characterizing abnormal data segments in the data set to be processed. The disclosed embodiment uses a data evaluation model to quickly and accurately monitor the abnormal state of data generated during software operation, which is conducive to realizing efficient software health evaluation.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method and device for generating a data evaluation model, a method and device for monitoring abnormal data, a computer-readable storage medium, and an electronic device. Background Art

[0002] The evaluation method of software project health is a very important research field in software engineering. A good software project health evaluation method can effectively conduct a systematic and comprehensive quality evaluation of software projects, provide necessary quality references for project monitoring personnel and software acceptance personnel, and provide a way for software developers to promptly discover problems and continuously improve software development practices, ultimately ensuring project success. Evaluating the data generated during software runtime is a common means of determining software health. At present, the architecture and operation process of many software are complex, with many driver modules, and high requirements for the management of software health and reliability. Therefore, a reliable and convenient method is needed to monitor the abnormality of data generated during software runtime. Summary of the invention

[0003] Embodiments of the present disclosure provide a data evaluation model generation method, device, computer-readable storage medium, and electronic device.

[0004] An embodiment of the present disclosure provides a method for generating a data evaluation model, the method comprising: obtaining a training sample, wherein the training sample comprises a sample data set and abnormal data annotation information for characterizing whether the sample data set includes an abnormal data segment; performing abnormal data determination on the sample data set through a preset model to be trained to obtain abnormal data prediction information; based on the abnormal data prediction information and the abnormal data annotation information, using a preset loss function, determining a loss value representing an error between the abnormal data prediction information and the abnormal data annotation information; based on the loss value, adjusting parameters of the model to be trained; in response to the model to be trained after adjusting parameters meeting a preset training end condition, determining the model to be trained after adjusting parameters as a data evaluation model.

[0005] According to another aspect of an embodiment of the present disclosure, a method for monitoring abnormal data is provided, the method comprising: obtaining a data set to be processed from a data storage area, wherein the data storage area is used to store data generated by the target software during operation; performing serialization preprocessing on the data set to be processed to obtain a serialized data set; and processing the serialized data set based on a pre-trained data evaluation model to obtain descriptive information for characterizing abnormal data segments in the data set to be processed.

[0006] According to another aspect of an embodiment of the present disclosure, a data evaluation model generating device is provided, the device comprising: a first acquisition module, used to acquire training samples, wherein the training samples include a sample data set and abnormal data annotation information used to characterize whether the sample data set includes abnormal data segments; a judgment module, used to perform abnormal data judgment on the sample data set through a preset model to be trained to obtain abnormal data prediction information; a first determination module, used to determine a loss value representing the error between the abnormal data prediction information and the abnormal data annotation information based on the abnormal data prediction information and the abnormal data annotation information and using a preset loss function; an adjustment module, used to adjust the parameters of the model to be trained based on the loss value; and a second determination module, used to determine the model to be trained after the parameters are adjusted as the data evaluation model in response to the model to be trained after the parameters are adjusted meeting a preset training end condition.

[0007] According to another aspect of an embodiment of the present disclosure, an abnormal data monitoring device is provided, which includes: a second acquisition module, used to acquire a data set to be processed from a data storage area, wherein the data storage area is used to store data generated by the target software during operation; a first processing module, used to perform serialization preprocessing on the data set to be processed to obtain a serialized data set; and a second processing module, used to process the serialized data set based on a pre-trained data evaluation model to obtain descriptive information for characterizing abnormal data segments in the data set to be processed.

[0008] According to another aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, which stores a computer program for executing the above-mentioned data evaluation model generating method or abnormal data monitoring method.

[0009] According to another aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a processor; a memory for storing instructions executable by the processor; and a processor for reading executable instructions from the memory and executing the instructions to implement the above-mentioned data evaluation model generating method or abnormal data monitoring method.

[0010] Based on the data evaluation model generation method and device, abnormal data monitoring method and device, computer-readable storage medium and electronic device provided by the above-mentioned embodiments of the present disclosure, the preset model to be trained is trained by using the sample data set and the corresponding abnormal data annotation information to obtain the data evaluation model. When monitoring the abnormal data of the software, the data to be processed generated by the software during operation is obtained, and the data to be processed is serialized, and then the serialized data set is processed using the trained data evaluation model, and finally the description information for characterizing the abnormal data segments in the data set to be processed is obtained. It realizes the rapid and accurate monitoring of the abnormal state of the data generated during the software operation, which is conducive to the efficient evaluation of the health of the software.

[0011] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. In the drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 is a system diagram to which the present disclosure is applicable.

[0014] Figure 2 It is a flowchart of a method for generating a data evaluation model provided by an exemplary embodiment of the present disclosure.

[0015] Figure 3 It is a flowchart of a method for generating a data evaluation model provided by an exemplary embodiment of the present disclosure.

[0016] Figure 4 It is a flowchart of a method for generating a data evaluation model provided by an exemplary embodiment of the present disclosure.

[0017] Figure 5 It is a flowchart of a method for generating a data evaluation model provided by an exemplary embodiment of the present disclosure.

[0018] Fig. 6A and Figure 6B is an exemplary schematic diagram of a visualization image provided by an exemplary embodiment of the present disclosure.

[0019] Figure 7 It is a flowchart of an abnormal data monitoring method provided by an exemplary embodiment of the present disclosure.

[0020] Figure 8 It is a structural diagram of a data evaluation model generating device provided by an exemplary embodiment of the present disclosure.

[0021] Fig. 9 It is a structural diagram of a data evaluation model generating device provided by another exemplary embodiment of the present disclosure.

[0022] Fig.10 It is a structural diagram of an abnormal data monitoring device provided by another exemplary embodiment of the present disclosure.

[0023] Fig.11 is a structural diagram of an electronic device provided by an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present disclosure will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited to the exemplary embodiments described here.

[0025] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless specifically stated otherwise.

[0026] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present disclosure are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.

[0027] It should also be understood that in the embodiments of the present disclosure, “plurality” may refer to two or more than two, and “at least one” may refer to one, two, or more than two.

[0028] It should also be understood that any component, data or structure mentioned in the embodiments of the present disclosure can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0029] In addition, the term "and / or" in the present disclosure is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are in an "or" relationship.

[0030] It should also be understood that the description of the various embodiments in the present disclosure focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.

[0031] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0032] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0033] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0034] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0035] The embodiments of the present disclosure can be applied to various types of processors, terminal devices, computer systems, servers and other electronic devices, which can operate with many other general or special computing system environments or configurations. Examples of well-known processors, terminal devices, computing systems, environments and / or configurations suitable for use with processors, terminal devices, computer systems, servers and other electronic devices include, but are not limited to: single-core processors, multi-core heterogeneous processors, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems and distributed cloud computing technology environments including any of the above systems, etc.

[0036] Electronic devices such as processors, terminal devices, computer systems, servers, etc. can be described in the general context of computer system executable instructions (such as program modules) executed by computer systems. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0037] Application Overview

[0038] Many current software have complex architectures and operating processes, with numerous driver modules, and high requirements for software health, performance, and reliability management. During software operation, it is often necessary to wait until a fault occurs before it can be detected, and it is impossible to monitor the software's operating status in advance and discover faults in advance. In particular, during the long-term operation of embedded products, hardware aging and other problems can also cause random software failures. Since it is impossible to monitor the software's operating status in advance, it is not conducive to determining the software's fault point.

[0039] To solve the above problems, the embodiment of the present disclosure trains a data evaluation model based on a deep learning method, which can be used to monitor the data generated during software operation in real time and determine data anomalies.

[0040] Exemplary Systems

[0041] Figure 1An exemplary system architecture 100 is shown to which the data evaluation model generation method or apparatus and the abnormal data monitoring method or apparatus according to the embodiments of the present disclosure can be applied.

[0042] like Figure 1 As shown, the system architecture 100 may include at least one processing unit 101 and a data storage area 102. Among them, any processing unit in the at least one processing unit 101 may execute the above-mentioned abnormal data monitoring method, or be provided with the above-mentioned abnormal data monitoring device to monitor the data generated by the software running on the processing unit or other processing units. The data evaluation model may be set in the processing unit that executes the abnormal data monitoring method, or may be set in other processing units. The above-mentioned at least one processing unit 101 may include a neural network acceleration unit dedicated to executing the data evaluation model.

[0043] Optionally, the at least one processing unit 101 may include at least one first processing unit and / or at least one second processing unit, that is, at least one processing unit 101 is arranged in a SOC (System on Chip). Among them, the first processing unit is usually a high-performance processing unit, such as an ARM cortex-a processing unit. The second processing unit is usually a lower-performance processing unit, such as an ARM cortex-m processing unit, a DSP processing unit, a risc-v processing unit, an ARM cortex-r processing unit, etc. The second processing unit is used to assume functions such as functional safety, real-time computing, peripheral control, and data processing. The above-mentioned abnormal data monitoring method can be executed by any first processing unit to monitor the software running on the second processing unit or other first processing units. The above-mentioned abnormal data monitoring method can also be executed by any second processing unit to monitor the software running on the first processing unit or other second processing units. The above-mentioned system on chip may also include a neural network acceleration unit for running a data evaluation model separately.

[0044] The data evaluation model generation method can be executed by any processing unit in the at least one processing unit 101, and accordingly, the data evaluation model generation device can be set in any processing unit in the at least one processing unit 101. The data evaluation model generation method can also be executed by other electronic devices, and accordingly, the data evaluation model generation device can be set in other electronic devices. When the data evaluation model generation method is executed by other electronic devices, the trained data evaluation model can be set to any processing unit in the at least one processing unit 101 through a communication connection.

[0045] The data storage area 102 is used to store data generated by the target software when it is running, and the target software is the monitored software. The data storage area 102 can be set in the same system on chip as the at least one processing unit 101, or can be set in other electronic devices.

[0046] It should be noted that Figure 1 The system architecture shown is only exemplary, and the above-mentioned data evaluation model generation method or abnormal data monitoring method can be performed by any type of electronic device, and accordingly, the data evaluation model generation device or abnormal data monitoring device can be set in any type of electronic device. The above-mentioned electronic device may include a server, and may also include any type of terminal device. For example, it includes but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc.

[0047] The server may be a model training server for training a data evaluation model or an abnormal data monitoring server for running an abnormal data monitoring method.

[0048] Exemplary Methods

[0049] Figure 2 FIG. 1 is a flow chart of a method for generating a data evaluation model provided by an exemplary embodiment of the present disclosure. Figure 1 On any processing unit or other electronic device) of at least one processing unit shown in Figure 2 As shown, the method comprises the following steps:

[0050] Step 201, obtaining training samples.

[0051] In this embodiment, the electronic device may obtain training samples locally or remotely, wherein the training samples include a sample data set and abnormal data annotation information used to indicate whether the sample data set includes abnormal data segments.

[0052] The sample data set may be a data set generated by a preset software when it is running. When the sample data set does not include abnormal data, the abnormal data annotation information may indicate that the sample data set does not include abnormal data, for example, the number "0" indicates that there is no abnormal data. When the sample data set includes abnormal data, the abnormal data annotation information may include but is not limited to at least one of the following: the address of the abnormal data segment, the size of the space occupied by the abnormal data segment, the specific value included in the abnormal data segment, etc.

[0053] Step 202: abnormal data determination is performed on the sample data set using a preset model to be trained to obtain abnormal data prediction information.

[0054] In this embodiment, the electronic device can perform abnormal data determination on the sample data set through a preset model to be trained to obtain abnormal data prediction information. The model to be trained can be a deep learning model, which can be constructed by CNN (Convolutional Neural Network), which can include convolutional layers, pooling layers, fully connected layers, classifiers, etc.

[0055] The sample data set can be input into the model to be trained, and the model to be trained can extract feature data of the sample data set, detect and classify the feature data, and output abnormal data prediction information. The abnormal data prediction information may include the position of the abnormal data segment in the sample data set (which may correspond to the storage address of the data) and the confidence level indicating that the abnormal data segment is abnormal data.

[0056] Step 203: Based on the abnormal data prediction information and the abnormal data labeling information, a loss value representing the error between the abnormal data prediction information and the abnormal data labeling information is determined using a preset loss function.

[0057] In this embodiment, the electronic device may determine a loss value representing an error between the abnormal data prediction information and the abnormal data labeling information based on the abnormal data prediction information and the abnormal data labeling information using a preset loss function.

[0058] The loss function may be a loss function for classification, such as a cross entropy loss function. The loss function is used to calculate the error between the abnormal data prediction information actually obtained and the abnormal data labeling information, and the error is represented by a loss value.

[0059] Step 204: Adjust the parameters of the model to be trained based on the loss value.

[0060] In this embodiment, the electronic device can adjust the parameters of the model to be trained based on the loss value. The process of adjusting the parameters of the model to be trained is the training process.

[0061] The training process of the model is an optimal solution solving process, in which the optimal solution is given by data annotation, that is, abnormal data annotation information in this embodiment. The process of fitting the model to the optimal solution is mainly iteratively carried out by minimizing the error. For a set of input sample data sets, a loss function is set, which can calculate the gap between the actual output of the model (i.e., abnormal data prediction information) and the expected output (i.e., abnormal data annotation information), and transmit the gap to the connection between each neuron in the neural network through the back propagation algorithm. The difference signal transmitted to each connection represents the contribution rate of the connection to the overall error. Then the original model parameters (also called weights) can be updated and modified using the gradient descent algorithm.

[0062] By repeatedly executing step 201 to step 204, that is, using multiple groups of training samples to iteratively train the model, the model after each iterative training is the model to be trained for the next training.

[0063] Step 205, in response to the model to be trained after the parameters are adjusted meeting the preset training end condition, the model to be trained after the parameters are adjusted is determined as the data evaluation model.

[0064] In this embodiment, the electronic device may determine the model to be trained after adjusting the parameters as the data evaluation model in response to the model to be trained after adjusting the parameters meeting a preset training end condition.

[0065] Specifically, the model is iteratively trained by repeatedly executing steps 201 to 204 using multiple sets of training samples, and after each training, it is determined whether the current model meets the training end condition. When the training end condition is met, the model after the current parameter adjustment is the trained data evaluation model. The training end condition may include but is not limited to at least one of the following: the loss value of the above loss function converges, the training time exceeds the preset time, and the number of training times exceeds the preset number of times.

[0066] The method provided by the above-mentioned embodiments of the present disclosure, by utilizing a sample data set and corresponding abnormal data annotation information, trains a preset model to be trained to obtain a data evaluation model, and can introduce the characteristics of real abnormal data into the model training process, so that the trained data evaluation model can evaluate the data set more accurately.

[0067] In some optional implementations, such as Figure 3 As shown, the above step 201 may include the following sub-steps:

[0068] Step 2011, obtaining positive training samples.

[0069] The positive training sample includes a positive sample data set and first abnormal data annotation information for characterizing that the positive sample data set does not include an abnormal data segment. As an example, the first abnormal data annotation information may include location information indicating a data segment in the positive sample data set and a confidence level (e.g., a number "0") indicating that the data segment is an abnormal data segment.

[0070] Step 2012, obtaining negative training samples.

[0071] The negative training sample includes a negative sample data set and second abnormal data annotation information for characterizing that the negative sample data set includes an abnormal data segment. As an example, the second abnormal data annotation information may include position information of the abnormal data segment in the negative sample data set and a confidence level indicating that the data indicated by the position information is abnormal data, such as a number "1".

[0072] During training, the model to be trained can be trained alternately using positive training samples and negative training samples. In actual training, there are usually multiple groups of positive training samples and negative training samples. Iterative training is performed by repeatedly performing steps 201 to 204 using multiple groups of positive training samples and negative training samples, and performing step 205 when the training end condition is met, to finally obtain a data evaluation model.

[0073] This embodiment obtains positive training samples and negative training samples for training, and can more specifically use different types of sample data sets to adjust model parameters, thereby improving the accuracy of data evaluation by the data evaluation model.

[0074] In some optional implementations, such as Figure 4 As shown, step 2011 includes the following sub-steps:

[0075] Step 20111, obtaining a data set generated when the first target software runs in a normal state as a normal initial data set.

[0076] The first target software may be software running locally or running on other electronic devices. Generally, for a software, its corresponding storage area in the memory may include a code segment and a data segment, and a normal initial data set may be obtained from the data segment.

[0077] Step 20112, perform serialization preprocessing on the normal initial data set, and determine the obtained serialized data set as the positive sample data set.

[0078] Among them, serialization preprocessing can be performed in various existing ways, and its purpose is to transform a large amount of data into data that can be processed by the model according to certain rules. For example, serialization preprocessing can include but is not limited to at least one of the following: normalization processing, determining the region of interest (ROI) in the positive sample data set, etc.

[0079] Step 20113, convert the positive sample data set into a visual image.

[0080] Specifically, the positive sample data set can be converted into a visualized image using an existing data visualization method. The visualized image can usually be a grayscale image. As an example, the grayscale value corresponding to each positive sample data in the positive sample data set can be determined according to a preset correspondence between the specific numerical value of the positive sample data and the grayscale value, and then a grayscale image can be generated based on the grayscale value corresponding to each positive sample data.

[0081] like Fig. 6A As shown, the grayscale distribution of the visualization image corresponding to the positive sample data set is relatively uniform.

[0082] Step 20114: annotate the positive sample data set based on the visualized image to generate first abnormal data annotation information corresponding to the positive sample data set.

[0083] The first abnormal data annotation information may include the location information of the annotated data segment in the positive sample data set and the confidence level (i.e., "0") that the annotated data segment is an abnormal data segment. Fig. 6A In the grayscale image shown, a position is marked with a rectangular box, which corresponds to the position of the normal data segment in the positive sample data set, and the corresponding confidence is set.

[0084] Step 20115: obtain positive training samples based on the positive sample data set and the first abnormal data annotation information.

[0085] Specifically, a correspondence between the first abnormal data labeling information and the positive sample data set may be established, and the first abnormal data labeling information and the positive sample data set may be determined as positive training samples.

[0086] This embodiment obtains a normal initial data set, generates a visual image and annotates it, so that normal data segments can be annotated more intuitively and quickly, thereby improving the efficiency of annotation. At the same time, the positive training samples reflect the characteristics of the data generated when the software is running normally, which helps to combine with the negative training samples to improve the recognition accuracy of the trained model.

[0087] In some optional implementations, such as Figure 5 As shown, step 2012 includes the following sub-steps:

[0088] Step 20121, obtaining a data set generated when the second target software runs in an abnormal state as an abnormal initial data set.

[0089] The second target software may be software running locally or running on other electronic devices. The second target software may be the same as or different from the first target software.

[0090] Step 20122, perform serialization preprocessing on the abnormal initial data set, and determine the obtained serialized data set as a negative sample data set.

[0091] The serialization preprocessing may be the same as the serialization preprocessing in step 20112 above.

[0092] Step 20123, convert the negative sample dataset into a visualization image.

[0093] The method for generating the visualization image may be the same as step 20113 above. Figure 6B As shown, the grayscale distribution of the grayscale image corresponding to the negative sample data set is uneven, with Figure 6B The rectangular box shown marks the mutation region.

[0094] Step 20124: Generate second abnormal data annotation information corresponding to the negative sample data set based on annotating the area representing the abnormal data segment in the visualized image.

[0095] The second abnormal data labeling information may include the position information of the labeled data segment in the negative sample data set and the confidence level (ie, the number “1”) that the labeled data segment is an abnormal data segment.

[0096] like Figure 6B As shown, the position of the rectangular box in the visualization image represents the position of the abnormal data segment, and the corresponding confidence level is set.

[0097] Step 20125: Obtain negative training samples based on the negative sample data set and the second abnormal data annotation information.

[0098] Specifically, a correspondence between the second abnormal data labeling information and the negative sample data set may be established, and the second abnormal data labeling information and the negative sample data set may be determined as positive training samples.

[0099] This embodiment obtains a normal initial data set, serializes the abnormal initial data set, generates a visual image and annotates it, so as to more intuitively and quickly determine the location of the abnormal data segment and improve the efficiency of annotation. At the same time, the negative training samples reflect the characteristics of the data generated when the software is running abnormally, which helps to combine with the positive training samples and improve the recognition accuracy of the trained model.

[0100] Figure 7 FIG. 1 is a flow chart of an abnormal data monitoring method provided by an exemplary embodiment of the present disclosure. Figure 1 On any processing unit of at least one processing unit shown in Figure 2 As shown, the method comprises the following steps:

[0101] Step 701: Obtain a data set to be processed from a data storage area.

[0102] In this embodiment, the electronic device can store data from a data storage area (such as Figure 1 The data storage area 102 shown in the figure obtains the data set to be processed. The data storage area is used to store the data generated by the target software when it is running. Usually, the data storage area is set in the memory, and the storage area corresponding to the target software in the memory may include a code segment and a data segment (i.e., the data storage area). The electronic device can determine the location of the data segment from the memory and extract the data set to be processed from the data segment.

[0103] Step 702: perform serialization preprocessing on the data set to be processed to obtain a serialized data set.

[0104] In this embodiment, the electronic device may perform serialization preprocessing on the data set to be processed to obtain a serialized data set. The serialization preprocessing may be performed in various existing ways, and its purpose is to transform a large amount of data into data that can be processed by the model according to certain rules. For example, the serialization preprocessing may include but is not limited to at least one of the following: normalization processing, determining a region of interest (ROI) in a positive sample data set, etc.

[0105] Step 703: Process the serialized data set based on the pre-trained data evaluation model to obtain description information for characterizing abnormal data segments in the data set to be processed.

[0106] In this embodiment, the electronic device can process the serialized data set based on the pre-trained data evaluation model to obtain description information for characterizing the abnormal data segments in the data set to be processed. Figure 2 The method provided in the corresponding embodiment is trained.

[0107] The serialized data set may be input into a data evaluation model, which processes the serialized data set and outputs description information, which may indicate whether the to-be-processed data set includes an abnormal data segment.

[0108] The method provided by the above-mentioned embodiment of the present disclosure obtains the data to be processed generated by the software during operation when monitoring abnormal data of the target software, performs serialization processing on the data to be processed, and then uses the trained data evaluation model to process the serialized data set, and finally obtains the description information for characterizing the abnormal data segments in the data set to be processed. The abnormal state monitoring of the data generated by the software during operation is realized quickly and accurately by using the data evaluation model, which is conducive to realizing efficient software health evaluation.

[0109] In some optional implementations, the description information includes at least one of the following: location information of the abnormal data segment in the data set to be processed and the confidence level of the abnormal data segment.

[0110] Generally, when the confidence level corresponding to the location information output by the data evaluation model is greater than or equal to a preset confidence threshold, the output description information may include information indicating that the data segment corresponding to the location information is abnormal data (e.g., in the form of text, symbols, etc.). When the confidence level corresponding to the location information output by the data evaluation model is less than the confidence threshold, the output description information may include information indicating that the data segment corresponding to the location information is normal data.

[0111] By outputting the position information, the abnormal data segment can be quickly located. By outputting the confidence level, the abnormal degree of the abnormal data segment can be described, providing a basis for users to perform abnormal data analysis.

[0112] In some optional implementations, the target software includes at least one of the following: software running locally and software running on the target processing unit. The software running locally is the electronic device that executes the abnormal data monitoring method, for example Figure 1 When a processing unit in the at least one processing unit 101 shown executes the abnormal data monitoring method, the target software may be software running in the processing unit, and the target processing unit is other processing units other than the processing unit. When the target processing unit is not other processing units executing the abnormal data monitoring method, the processing unit for executing the abnormal data monitoring method may obtain the data set to be processed from a pre-specified data storage area corresponding to the target software.

[0113] Optionally, the at least one processing unit 101 may include at least one first processing unit and / or at least one second processing unit, that is, at least one processing unit 101 is arranged in the SOC. The first processing unit is usually a high-performance processing unit, such as an ARM cortex-a processing unit. The second processing unit is usually a lower-performance processing unit, such as an ARM cortex-m processing unit, a DSP processing unit, a risc-v processing unit, an ARM cortex-r processing unit, etc. The second processing unit is used to assume functions such as functional safety, real-time computing, peripheral control, and data processing.

[0114] The abnormal data monitoring method can be executed by any first processing unit, and the target software can be software running in the first processing unit, or software running in the second processing unit or other first processing units. The abnormal data monitoring method can be executed by any second processing unit, and the target software can be software running in the second processing unit, or software running in the first processing unit or other second processing units.

[0115] This embodiment can flexibly configure the monitored software by monitoring abnormal data of the target software running locally or in the target processing unit. When the target software is software running in the target processing unit, the data processing volume of the target processing unit can be reduced and the operating efficiency of the target processing unit can be improved.

[0116] Optionally, when executing step 703, the serialized data set may be input into a neural network acceleration unit, and the neural network acceleration unit is used to run the data evaluation model. The neural network acceleration unit may be as follows: Figure 1 A processing unit in at least one processing unit 101 shown. By setting the data evaluation model in the neural network acceleration unit, the data processing amount of the electronic device executing the abnormal data monitoring method can be reduced, and the efficiency of abnormal data monitoring can be improved.

[0117] Exemplary Devices

[0118] Figure 8 is a schematic diagram of the structure of a data evaluation model generation device provided by an exemplary embodiment of the present disclosure. This embodiment can be applied to electronic devices, such as Figure 8As shown, the data evaluation model generating device includes: a first acquisition module 801, which is used to acquire training samples, wherein the training samples include a sample data set and abnormal data annotation information used to characterize whether the sample data set includes abnormal data segments; a judgment module 802, which is used to perform abnormal data judgment on the sample data set through a preset model to be trained to obtain abnormal data prediction information; a first determination module 803, which is used to determine a loss value representing the error between the abnormal data prediction information and the abnormal data annotation information based on the abnormal data prediction information and the abnormal data annotation information, using a preset loss function; an adjustment module 804, which is used to adjust the parameters of the model to be trained based on the loss value; and a second determination module 805, which is used to determine the model to be trained after the parameters are adjusted as the data evaluation model in response to the model to be trained after the parameters are adjusted meeting the preset training end condition.

[0119] In this embodiment, the first acquisition module 801 may acquire training samples locally or remotely, wherein the training samples include a sample data set and abnormal data annotation information used to indicate whether the sample data set includes abnormal data segments.

[0120] The sample data set may be a data set generated by the preset software when it is running. When the sample data set does not include abnormal data, the abnormal data annotation information may indicate that the sample data set does not include abnormal data, for example, the number "0" indicates that there is no abnormal data. When the sample data set includes abnormal data, the abnormal data annotation information may include but is not limited to at least one of the following: the address of the abnormal data segment, the size of the space occupied by the abnormal data segment, the specific value included in the abnormal data segment, etc.

[0121] In this embodiment, the determination module 802 can perform abnormal data determination on the sample data set through a preset model to be trained to obtain abnormal data prediction information. The model to be trained can be a deep learning model, which can be constructed by CNN (Convolutional Neural Network), which can include convolutional layers, pooling layers, fully connected layers, classifiers, etc.

[0122] The sample data set can be input into the model to be trained, and the model to be trained can extract feature data of the sample data set, detect and classify the feature data, and output abnormal data prediction information. The abnormal data prediction information may include the position of the abnormal data segment in the sample data set (which may correspond to the storage address of the data) and the confidence level indicating that the abnormal data segment is abnormal data.

[0123] In this embodiment, the first determination module 803 may determine a loss value representing an error between the abnormal data prediction information and the abnormal data annotation information based on the abnormal data prediction information and the abnormal data annotation information using a preset loss function.

[0124] The loss function may be a loss function for classification, such as a cross entropy loss function. The loss function is used to calculate the error between the abnormal data prediction information actually obtained and the abnormal data labeling information, and the error is represented by a loss value.

[0125] In this embodiment, the adjustment module 804 can adjust the parameters of the model to be trained based on the loss value. The process of adjusting the parameters of the model to be trained is the training process.

[0126] The training process of the model is an optimal solution solving process, in which the optimal solution is given by data annotation, that is, abnormal data annotation information in this embodiment. The process of fitting the model to the optimal solution is mainly iteratively carried out by minimizing the error. For a set of input sample data sets, a loss function is set, which can calculate the gap between the actual output of the model (i.e., abnormal data prediction information) and the expected output (i.e., abnormal data annotation information), and transmit the gap to the connection between each neuron in the neural network through the back propagation algorithm. The difference signal transmitted to each connection represents the contribution rate of the connection to the overall error. Then the original model parameters (also called weights) can be updated and modified using the gradient descent algorithm.

[0127] By repeatedly executing step 201 to step 204, that is, using multiple groups of training samples to iteratively train the model, the model after each iterative training is the model to be trained for the next training.

[0128] In this embodiment, the second determination module 805 may determine the model to be trained after adjusting the parameters as the data evaluation model in response to the model to be trained after adjusting the parameters meeting a preset training end condition.

[0129] Specifically, the model is iteratively trained by repeatedly executing steps 201 to 204 using multiple sets of training samples, and after each training, it is determined whether the current model meets the training end condition. When the training end condition is met, the model after the current parameter adjustment is the trained data evaluation model. The training end condition may include but is not limited to at least one of the following: the loss value of the above loss function converges, the training time exceeds the preset time, and the number of training times exceeds the preset number of times.

[0130] Reference Fig. 9 , Fig. 9 It is a structural diagram of a data evaluation model generating device provided by another exemplary embodiment of the present disclosure.

[0131] In some optional implementations, the first acquisition module 801 includes: a first acquisition unit 8011, used to acquire positive training samples, wherein the positive training samples include a positive sample data set and first abnormal data annotation information for characterizing that the positive sample data set does not include abnormal data segments; a second acquisition unit 8012, used to acquire negative training samples, wherein the negative training samples include a negative sample data set and second abnormal data annotation information for characterizing that the negative sample data set includes abnormal data segments.

[0132] In some optional implementations, the first acquisition unit 8011 includes: a first acquisition sub-unit 80111, used to acquire a data set generated when the first target software runs in a normal state as a normal initial data set; a first preprocessing sub-unit 80112, used to perform serialization preprocessing on the normal initial data set, and determine the obtained serialized data set as a positive sample data set; a first conversion sub-unit 80113, used to convert the positive sample data set into a visual image; a first annotation sub-unit 80114, used to annotate the positive sample data set based on the visual image, and generate first abnormal data annotation information corresponding to the positive sample data set; a first generation sub-unit 80115, used to obtain a positive training sample based on the positive sample data set and the first abnormal data annotation information.

[0133] In some optional implementations, the second acquisition unit 8012 includes: a second acquisition sub-unit 80121, used to acquire a data set generated when the second target software runs in an abnormal state as an abnormal initial data set; a second preprocessing sub-unit 80122, used to serialize and preprocess the abnormal initial data set, and determine the obtained serialized data set as a negative sample data set; a second conversion sub-unit 80123, used to convert the negative sample data set into a visual image; a second annotation sub-unit 80124, used to generate second abnormal data annotation information corresponding to the negative sample data set based on annotating the area representing the abnormal data segment in the visual image; a second generation sub-unit 80125, used to obtain a negative training sample based on the negative sample data set and the second abnormal data annotation information.

[0134] The data evaluation model generating device provided in the above-mentioned embodiment of the present disclosure trains a preset model to be trained by utilizing a sample data set and corresponding abnormal data annotation information to obtain a data evaluation model. This allows the model training process to introduce features of real abnormal data, so that the trained data evaluation model can evaluate the data set more accurately.

[0135] Fig.10 is a schematic diagram of the structure of an abnormal data monitoring device provided by an exemplary embodiment of the present disclosure. This embodiment can be applied to electronic devices, such as Fig.10As shown, the abnormal data monitoring device includes: a second acquisition module 1001, used to obtain a data set to be processed from a data storage area, wherein the data storage area is used to store data generated by the target software during operation; a first processing module 1002, used to perform serialization preprocessing on the data set to be processed to obtain a serialized data set; a second processing module 1003, used to process the serialized data set based on a pre-trained data evaluation model to obtain descriptive information for characterizing abnormal data segments in the data set to be processed.

[0136] In this embodiment, the second acquisition module 1001 can store data in a data storage area (such as Figure 1 The data storage area 102 shown in the figure obtains the data set to be processed. The data storage area is used to store the data generated by the target software when it is running. Usually, the data storage area is set in the memory, and the storage area corresponding to the target software in the memory may include a code segment and a data segment (i.e., the data storage area). The electronic device can determine the location of the data segment from the memory and extract the data set to be processed from the data segment.

[0137] In this embodiment, the first processing module 1002 can perform serialization preprocessing on the data set to be processed to obtain a serialized data set. The serialization preprocessing can be performed in various existing ways, and its purpose is to transform a large amount of data into data that can be processed by the model according to certain rules. For example, the serialization preprocessing may include but is not limited to at least one of the following: normalization processing, determining a region of interest (ROI) in a positive sample data set, etc.

[0138] In this embodiment, the second processing module 1003 can process the serialized data set based on the pre-trained data evaluation model to obtain description information for characterizing the abnormal data segments in the data set to be processed. Figure 2 The method provided in the corresponding embodiment is trained.

[0139] The serialized data set may be input into a data evaluation model, which processes the serialized data set and outputs description information, which may indicate whether the to-be-processed data set includes an abnormal data segment.

[0140] In some optional implementations, the target software includes at least one of the following: software running locally and software running on a target processing unit.

[0141] In some optional implementations, the description information includes at least one of the following: location information of the abnormal data segment in the data set to be processed and the confidence level of the abnormal data segment.

[0142] The abnormal data monitoring device provided by the above embodiment of the present disclosure obtains the data to be processed generated by the software when it is running when monitoring the abnormal data of the target software, and performs serialization processing on the data to be processed, and then uses the trained data evaluation model to process the serialized data set, and finally obtains the description information for characterizing the abnormal data segment in the data set to be processed. It realizes the abnormal state monitoring of the data generated by the software when it is running quickly and accurately by using the data evaluation model, which is conducive to realizing the efficient evaluation of the health of the software.

[0143] Exemplary Electronic Devices

[0144] Below, reference Fig.11 An electronic device according to an embodiment of the present disclosure is described. Fig.11 A block diagram of an electronic device according to an embodiment of the present disclosure is illustrated.

[0145] like Fig.11 As shown, the electronic device 1100 includes a monitoring processing unit 1101 , a monitored processing unit 1102 , and a memory 1103 for storing executable instructions of the monitoring processing unit 1101 and the monitored processing unit 1102 .

[0146] The monitoring processing unit 1101 and the monitored processing unit 1102 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 1100 to perform desired functions.

[0147] The monitored processing unit 1102 is used to run the target software to generate a data set to be processed.

[0148] The monitoring processing unit 1101 is used to read the corresponding executable instructions from the memory and execute the read instructions to use the data set to be processed as the sample data set to implement the above Figure 2 The data evaluation model generation method provided in the corresponding embodiment and its optional implementation mode, or the above-mentioned method is implemented based on the data set to be processed. Figure 7 The corresponding embodiment and its optional implementation method provide an abnormal data monitoring method.

[0149] The memory 1103 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on a computer-readable storage medium.

[0150] It should be noted that the number of monitoring processing units 1101 and the number of monitored processing units 1102 can be set arbitrarily. For example, the monitoring processing unit 1101 and the monitored processing unit 1102 can constitute a multi-core heterogeneous system on chip. The system on chip may include at least two processing units, and any processing unit of the at least two processing units can be configured as a monitoring processing unit 1101 or a monitored processing unit 1102. Optionally, the at least two processing units may include at least one first processing unit and at least one second processing unit. Among them, the first processing unit is usually a high-performance processing unit, such as an ARM cortex-a processing unit. The second processing unit is usually a lower-performance processing unit, such as an ARM cortex-m processing unit, a DSP processing unit, a risc-v processing unit, an ARM cortex-r processing unit, etc. The second processing unit is used to assume functions such as functional safety, real-time computing, peripheral control, and data processing.

[0151] As an example, any first processing unit among the at least one first processing unit may be set as the monitoring processing unit 1101, and any second processing unit among the at least one second processing unit may be set as the monitored processing unit 1102. Alternatively, any second processing unit among the at least one second processing unit may be set as the monitoring processing unit 1101, and any first processing unit among the at least one first processing unit may be set as the monitored processing unit 1102.

[0152] In one example, the electronic device 1100 may further include: an input device 1104 and an output device 1105 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0153] For example, the input device 1104 may be a device such as a mouse or a keyboard, which is used to input training samples, data to be processed, various commands required for executing the above method, etc. The input device 1104 may also be a communication network connector, which is used to receive training samples, data to be processed, various commands required for executing the above method, etc. from a remote location.

[0154] The output device 1105 can output various information to the outside, including system startup progress information, etc. The output device 1105 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0155] Of course, to simplify, Fig.11 Only some of the components related to the present disclosure in the electronic device 1100 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application scenarios, the electronic device 1100 may also include any other appropriate components.

[0156] Exemplary computer program products and computer-readable storage media

[0157] In addition to the above-mentioned methods and devices, an embodiment of the present disclosure may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the data evaluation model generation method or abnormal data monitoring method according to various embodiments of the present disclosure described in the above "Exemplary Method" section of this specification.

[0158] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the disclosed embodiments, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0159] In addition, an embodiment of the present disclosure may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the steps of the data evaluation model generation method or abnormal data monitoring method according to various embodiments of the present disclosure described in the above “Exemplary Method” section of this specification.

[0160] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0161] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0162] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0163] The block diagrams of the devices, apparatuses, equipment, and systems involved in this disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open words, referring to "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0164] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0165] It should also be noted that in the apparatus, device and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0166] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0167] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for generating a data evaluation model, comprising: Acquire a training sample, wherein the training sample includes a sample data set and abnormal data annotation information used to characterize whether the sample data set includes an abnormal data segment, and the abnormal data annotation information includes: an address of the abnormal data segment, a size of a space occupied by the abnormal data segment, and a specific value included in the abnormal data segment; By using a preset model to be trained, abnormal data determination is performed on the sample data set to obtain abnormal data prediction information, wherein the abnormal data prediction information includes position information of the abnormal data segment in the sample data set and a confidence level indicating that the abnormal data segment is abnormal data, and the position information is used to locate the abnormal data segment; Based on the abnormal data prediction information and the abnormal data labeling information, using a preset loss function, determining a loss value representing an error between the abnormal data prediction information and the abnormal data labeling information; Based on the loss value, adjusting the parameters of the model to be trained; In response to the model to be trained after the parameters are adjusted meeting a preset training end condition, the model to be trained after the parameters are adjusted is determined as a data evaluation model.

2. The method according to claim 1, wherein: The obtaining of training samples comprises: Acquire a positive training sample, wherein the positive training sample includes a positive sample data set and first abnormal data annotation information for characterizing that the positive sample data set does not include an abnormal data segment; A negative training sample is obtained, wherein the negative training sample includes a negative sample data set and second abnormal data annotation information for characterizing that the negative sample data set includes an abnormal data segment.

3. The method according to claim 2, wherein: The obtaining of positive training samples comprises: Acquire a data set generated when the first target software is running in a normal state as a normal initial data set; Performing serialization preprocessing on the normal initial data set, and determining the obtained serialized data set as a positive sample data set; Convert the positive sample data set into a visual image; Based on the visualized image, annotate the positive sample data set to generate first abnormal data annotation information corresponding to the positive sample data set; The positive training sample is obtained based on the positive sample data set and the first abnormal data labeling information.

4. The method according to claim 2, wherein: The obtaining of negative training samples comprises: Acquire a data set generated when the second target software is running in an abnormal state as an abnormal initial data set; Performing serialization preprocessing on the abnormal initial data set, and determining the obtained serialized data set as a negative sample data set; Convert the negative sample data set into a visual image; Generate second abnormal data annotation information corresponding to the negative sample data set based on annotating the area representing the abnormal data segment in the visualized image; The negative training samples are obtained based on the negative sample data set and the second abnormal data labeling information.

5. A method for monitoring abnormal data, comprising: Acquire a data set to be processed from a data storage area, wherein the data storage area is used to store data generated when the target software is running; Performing serialization preprocessing on the data set to be processed to obtain a serialized data set; The serialized data set is processed based on a pre-trained data evaluation model to obtain descriptive information for characterizing the abnormal data segments in the data set to be processed, wherein the descriptive information includes: position information of the abnormal data segments in the data set to be processed and the confidence of the abnormal data segments, and the position information is used to locate the abnormal data segments.

6. The method according to claim 5, wherein: The target software includes at least one of the following: software running locally and software running on a target processing unit.

7. The method according to claim 5 or 6, wherein: The description information includes at least one of the following: location information of the abnormal data segment in the data set to be processed, and confidence of the abnormal data segment.

8. A data evaluation model generating device, comprising: A first acquisition module is used to acquire a training sample, wherein the training sample includes a sample data set and abnormal data annotation information used to characterize whether the sample data set includes an abnormal data segment, and the abnormal data annotation information includes: an address of the abnormal data segment, a size of a space occupied by the abnormal data segment, and a specific value included in the abnormal data segment; A determination module, used to perform abnormal data determination on the sample data set through a preset model to be trained to obtain abnormal data prediction information, wherein the abnormal data prediction information includes position information of the abnormal data segment in the sample data set and a confidence level indicating that the abnormal data segment is abnormal data, and the position information is used to locate the abnormal data segment; A first determination module, configured to determine, based on the abnormal data prediction information and the abnormal data labeling information, a loss value representing an error between the abnormal data prediction information and the abnormal data labeling information using a preset loss function; An adjustment module, used for adjusting the parameters of the model to be trained based on the loss value; The second determination module is used to determine the model to be trained after the parameters are adjusted as the data evaluation model in response to the model to be trained after the parameters are adjusted meeting the preset training end condition.

9. An abnormal data monitoring device, comprising: A second acquisition module is used to acquire a data set to be processed from a data storage area, wherein the data storage area is used to store data generated when the target software is running; A first processing module, configured to perform serialization preprocessing on the data set to be processed to obtain a serialized data set; The second processing module is used to process the serialized data set based on a pre-trained data evaluation model to obtain descriptive information for characterizing the abnormal data segments in the data set to be processed, wherein the descriptive information includes: position information of the abnormal data segments in the data set to be processed and the confidence of the abnormal data segments, wherein the position is used to locate the abnormal data segments, and the confidence is used to describe the degree of abnormality of the abnormal data segments.

10. An electronic device, comprising a monitoring processing unit, a monitored processing unit, and a memory for storing executable instructions of the monitoring processing unit and the monitored processing unit; The monitored processing unit is used to run the target software to generate a data set to be processed; The monitoring processing unit is used to read the corresponding executable instructions from the memory and execute the read instructions to implement any of the methods described in claims 1-4 above using the data set to be processed as a sample data set, or to implement any of the methods described in claims 5-7 above based on the data set to be processed.

11. A computer-readable storage medium storing a computer program, wherein the computer program is used to execute the method according to any one of claims 1 to 7.

Images