CN110502445A - Software fault menace level determination method and device, model training method and device - Google Patents
Software fault menace level determination method and device, model training method and device Download PDFInfo
- Publication number
- CN110502445A CN110502445A CN201910805751.3A CN201910805751A CN110502445A CN 110502445 A CN110502445 A CN 110502445A CN 201910805751 A CN201910805751 A CN 201910805751A CN 110502445 A CN110502445 A CN 110502445A
- Authority
- CN
- China
- Prior art keywords
- fault
- attribute
- software
- software fault
- menace level
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/362—Software debugging
- G06F11/366—Software debugging using diagnostics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
- Stored Programmes (AREA)
Abstract
The embodiment of the invention provides a kind of software fault menace level determination method and devices, model training method and device.The determination method includes: to be predicted in real time using multiple breakdown judge models failure；The real-time prediction result of multiple breakdown judge models is weighted and averaged, the fault level judgement to software fault is obtained.Software fault menace level determination method and device, model training method and device provided in an embodiment of the present invention can guarantee that grade determines the data volume that model training data are substantially reduced under the premise of accuracy.
Description
Technical field
The present invention relates to fault detection technique fields, more particularly to a kind of software fault menace level determination method and dress
It sets, model training method and device.
Background technique
Software fault menace level determines the priority of fault restoration, if it is decided that fault is likely to result in serious
Failure is not repaired in time, to will lead to the serious consequence of software quality difference.Software fault menace level needs to rely at present
Manually determined, and manually determine the menace level of failure there are following both sides disadvantages:
The accuracy of the artificial menace level for determining failure relies primarily on the experience of people, and the personnel being lacking in experience will lead
Cause decision error.Different personnel are easy to have differences to the standard understanding that menace level determines, may cause different testers
Difference is determined to the menace level of same problem, influences the accuracy of speed measuring with software evaluation.
Since the number of faults of large-scale software is big, artificial judgement fault level low efficiency is relied on, will lead to number of faults
According to collecting slowly, the progress of software fault reparation will affect.
If judge using preset Fault Model the automatic judgement of menace level, it usually needs to carrying out etc.
The model that grade determines is trained.This training usually requires a certain amount of labeled data, as the training data being trained.
It will be apparent to those skilled in the art that needing to carry out attribute labeling to training data before being trained model.This mark
Note work is time-consuming and laborious, however, in order to guarantee that the accuracy of model output result has to carry out this laborious work.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of software fault menace level determination method and devices, model instruction
Practice method and device, can guarantee that grade determines the data volume for substantially reducing model training data under the premise of accuracy.
In order to solve the above technical problems, the present invention provides a kind of software fault menace level decision model training method,
The described method includes: according to the fault data of software fault, the multiple and different breakdown judge moulds based on maximum entropy theory of training
Type；According to the fault message of software fault, the multiple and different breakdown judge models based on maximum entropy theory of training, comprising: obtain
Take the fault attribute of various software faults；The fault attribute is converted into fault feature vector；Using the fault signature to
Amount, the multiple breakdown judge models based on maximum entropy theory of training.
In some embodiments, the different types of fault attribute got is endowed different attribute weight.
In some embodiments, the fault attribute for type of making a summary is endowed attribute weight 3, the failure of discovery time type
Attribute is endowed attribute weight 1, and the fault attribute of software version type is endowed attribute weight 1, describes the fault attribute of type
It is endowed attribute weight 2, the fault attribute of subject categories is endowed attribute weight 2, and the fault attribute of annotation category is endowed category
Property weight 1.
In some embodiments, occur identical word in different types of attribute to distinguish, will add before each word
Upper fault attribute name is as prefix.
In some embodiments, it is based on one-hot coding, the fault attribute is converted into fault feature vector, comprising:
Judge whether to need to be segmented and gone stop words to the fault attribute got；To the word sequence got by participle
Carry out one-hot coding.
In addition, the embodiment of the invention also provides a kind of software fault menace level determination methods, which comprises benefit
Failure is predicted in real time with multiple breakdown judge models based on maximum entropy theory；To multiple breakdown judge models
Real-time prediction result is weighted and averaged, and obtains the fault level judgement to software fault.
In some embodiments, the real-time prediction result of multiple breakdown judge models is weighted and averaged, is obtained
To the fault level judgement to software fault, comprising: be weighted and averaged according to following calculation formula, to obtain to software fault
Fault level judgement:
Wherein, wiIt is the weight of model i, hi(x) be model prediction result.
In some embodiments, the weight of model i learns to obtain by following formula:
Wherein, accuracy rate of the model i in test data set j is cij。
In addition, the present invention also provides a kind of software fault menace level decision model training device, described device includes:
One or more processors；Storage device, for storing one or more programs, when one or more of programs are by described one
A or multiple processors execute, so that one or more of processors are realized according to previously described software fault menace level
Decision model training method.
In addition, the present invention also provides a kind of software fault menace level decision maker, described device includes: one or more
A processor；Storage device, for storing one or more programs, when one or more of programs are one or more of
Processor executes, so that one or more of processors are realized according to previously described software fault menace level judgement side
Method.
By adopting such a design, the present invention has at least the following advantages:
1, the software fault data for not needing largely to have marked are as model training data:
Have the largely software fault data that have marked, be good to the training of model, though however reality be
A large amount of software fault data are so had collected, these data do not mark manually, this is because the cost manually marked is big.At this moment
It is required that model is for the model that has a small amount of labeled data that can also train.
For single model, training data tends not to obtain effective model less, the method that we use integrated study,
The multiple models of training, to avoid the few influence to model of training data.
2, failure menace level determines that accuracy height, decision metrics consistency are substantially improved:
Maximum entropy model can show good effect for limited training data, with Nae Bayesianmethod phase
Than its advantages are that do not have attribute conditions independence assumption, therefore select maximum entropy model, and failure menace level determines accuracy
It is high.
Based on maximum entropy model, increase by the way that the weight of software fault attribute is arranged, and to the fault data after participle
The prefix of fault attribute improves the training effect of maximum entropy model so that the input data of maximum entropy model is more reasonable.
In order to avoid single model be easy to cause the risk of over-fitting under limited training data, we use integrated study
Thought, the different fault attribute subset of option, the different model of training combines different moulds using effective combined strategy
Type prediction result, to obtain final result.
The cost of software fault menace level decision error is analyzed, for the prediction effect of effective assessment models, is used
Cost-sensitive error rate, so that the assessment of model is more in line with the actual conditions of application.
Detailed description of the invention
The above is merely an overview of the technical solutions of the present invention, in order to better understand the technical means of the present invention, below
In conjunction with attached drawing, the present invention is described in further detail with specific embodiment.
Fig. 1 is the flow diagram of software fault menace level decision model training method provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of software fault menace level determination method provided in an embodiment of the present invention；
Fig. 3 is the structural schematic diagram of software fault menace level decision maker provided in an embodiment of the present invention.
Specific embodiment
Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings, it should be understood that preferred reality described herein
Apply example only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
The method that software fault menace level decision technology uses machine learning, is based on maximum entropy model, and utilization is existing
Software fault data are trained, and obtain software fault menace level decision model by training, new using the model prediction
Software fault menace level.
(a), the software fault menace level decision model training based on maximum entropy model, is gone through using existing software fault
History data are based on maximum entropy model training software failure menace level decision model, two kinds are proposed in terms of the training of model
Improvement strategy, to improve the accuracy of model.
(b), the model completed has been trained in the software fault menace level decision model prediction based on maximum entropy model, utilization
It predicts the menace level of new software fault, and prediction result is assessed using assessment level.
Software fault menace level decision model training based on maximum entropy model:
Step 1, the attribute for selecting failure.
Description software fault information have very much, such as: project name, problem I D, problem state, abstract, discovery time,
Software version executes round, description, theme, annotation, if all attributes all used, first is that the time of training pattern is multiple
Miscellaneous degree is high, second is that the accuracy of model is low instead.
Table 1
Failure is converted feature vector by step 2.
Which attribute needs of analysis software fault first segment, and determine whether to segment the description side needed according to attribute
Formula and feature determine the universal criterious whether attribute segments:
If a) value of attribute can be enumerated, participle is not had to；
If b) attribute is description information, and can not be enumerated, then need to segment.
After participle, followed by remove stop words, stop words refers to some very universal vocabulary used, such as:, be
Deng these words do not have special meaning, are unable to the feature of fault data, useless for the menace level of failure.
After the above pretreated step, software fault information is converted to a word sequence, and word sequence cannot be direct
As the input of maximum entropy model, need each word in word sequence being converted to term vector, the method for use is one-hot coding.
One-hot coding is the most common most basic method that word is converted to vector, by one different integer of each word
Index is associated, integer index i is then converted to the binary vector (size that N is word finder) that size is N, which removes
Position i is 1, and other positions are all 0.
There is word frequency coding to the improved method of one-hot coding, word frequency coding is converted to the integer vectors that size is N for i is indexed
(size that N is word finder), for the vector in addition to the number that position i is that word occurs in a document, other positions are all 0.Word frequency-is inverse
Document frequency coding is the improvement to word frequency coding, and multiplied by inverse document frequency on the basis of word frequency coding, each word is converted
The real vector for being N for size.
Step 3, training pattern.
After word is converted to the identical term vector of dimension, the text information of software fault data is extracted, by the word in text
It is substituted for term vector, software fault data are just converted to vector by the addition of these term vectors, as the defeated of model training
Enter data.
The training of model is that the training method of machine learning model corresponding with machine learning model, different is different,
Therefore preference pattern is needed before model training.Include due to software fault data is text information, selectable classification
Model has: support vector machines, logistical regression, maximum entropy classification, gradient boosted tree etc..
By the experimental results showed that being better than other disaggregated models using maximum entropy model.Maximum entropy is probabilistic model study
A criterion, maximum entropy mode is to be generalized to classification problem.
The training of maximum entropy model can turn to constrained optimization problem in the form of, given training dataset T=(x1,
Y1), (x2, y2) ..., (xn, yn) } and characteristic function fi(x, y), i=1,2 ..., the study of n, maximum entropy model are of equal value
In constrained optimization problem:
The effect for the model that training obtains depends not only on the machine learning model of selection, also relies on trained input
Data, the judgement for failure menace level, due to being to extract word feature from fault attribute, it is clear that identical in different attribute
Word is different relative to the importance of failure menace level, such as: the word occurred in abstract is more important than the word occurred in annotation, because
This proposes two kinds of improvement strategies:
A) weight is increased to the word occurred in attribute important in failure, such as: higher to the word setting occurred in abstract
Weight；
B) there is identical word to be different from failure different attribute, will before each word plus fault attribute name as
Prefix, such as: occurring this word of "abnormal" in abstract, then convert " abstract: abnormal " for the word.
As shown in Figure 1, software fault menace level decision model training method provided by the invention includes: S11, obtain each
The fault attribute of kind software fault；The fault attribute is converted to fault feature vector by S12；S13, it is special using the failure
Levy vector, the multiple breakdown judge models based on maximum entropy theory of training.
The prediction effect of single model is often good without multiple models, the thought of random forest is used for reference, from fault attribute
Different attribute sets is chosen, the different model of attribute set training is utilized.
Occur identical word in software fault different attribute to distinguish, will before each word plus fault attribute name as
Prefix is the original creation of the application.Maximum entropy model is used for software fault menace level decision model, and uses different attributes
The different maximum entropy model of trained is the original creation of the application.
Software fault menace level decision model prediction based on maximum entropy model:
Using the thought of integrated study, as shown in Fig. 2, by combining multiple model prediction results as final
Output.
Key is combined strategy, we use weighted mean method:
It is wherein the weight of model i, is the prediction result of model, output is obtained by weighted average, weight is to pass through
Model i from the acquistion of test data middle school to, such as: model i is in the accuracy rate that test data is concentrated, then,
Consequence caused by the decision error of different type software fault menace level is different, such as menace level is high
Software fault mistake is determined as inferior grade, will lead to high-grade software fault and is not repaired in time, and consequence is software matter
It is high to measure risk；And the low software fault mistake of menace level being determined as high-grade, consequence is that the software fault of inferior grade obtains
It is preferential to repair, it wastes and exploits natural resources, will not be impacted to software quality.Therefore it needs to different types of mistake setting not
Same cost is called cost-sensitive error rate in machine learning field.
Using cost-sensitive error rate come assessment models prediction, first have to determine type of error, software fault it is serious etc.
Grade is generally divided into level Four, and one/secondary failure is known as high-grade failure by us, and three/level Four is inferior grade failure, type of error point
For two classes: Error type I be inferior grade failure by high-grade fault verification, error type II is by inferior grade fault verification
For high-grade failure.The cost of Error type I is expressed as cost1, the cost of error type II is expressed as cost2。cost1With
cost2Value be usually need to determine based on user demand.
The combined strategy of model is existing method, is used for the combination of maximum entropy model, is the original creation of the application.Generation
Valence sensitivity error rate is usually used in the disaggregated model of machine learning, and the menace level decision model for being used for assessment software fault is
The original creation of the application.
Fig. 3 is the structure chart of invention software failure menace level decision maker.Referring to Fig. 3, based on syntax conversion from
Dynamic test device includes: central processing unit (CPU) 301, can according to the program being stored in read-only memory (ROM) or
Person from the program that storage section 308 is loaded into random access storage device (RAM) 303 execute various movements appropriate and from
Reason.In RAM 303, it is also stored with various programs and data needed for system operatio.CPU 301, ROM 302 and RAM
303 are connected with each other by bus 304.Input/output (I/O) interface 305 is also connected to bus 304.
I/O interface 305 is connected to lower component: the importation 306 including keyboard, mouse etc.；It is penetrated including such as cathode
The output par, c 307 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.；Storage section 308 including hard disk etc.；
And the communications portion 309 of the network interface card including LAN card, modem etc..Communications portion 309 via such as because
The network of spy's net executes communication process.Driver 310 is also connected to I/O interface 305 as needed.Detachable media 311, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 310, in order to read from thereon
Computer program be mounted into storage section 308 as needed.
Similar, software fault menace level decision model training device provided by the invention also can have such as Fig. 3
Shown in structure.
The embodiment of the present invention obtain the utility model has the advantages that
1, the software fault data for not needing largely to have marked are as model training data:
Have the largely software fault data that have marked, be good to the training of model, though however reality be
A large amount of software fault data are so had collected, these data do not mark manually, this is because the cost manually marked is big.At this moment
It is required that model is for the model that has a small amount of labeled data that can also train.
For single model, training data tends not to obtain effective model less, the method that we use integrated study,
The multiple models of training, to avoid the few influence to model of training data.
2, failure menace level determines that accuracy height, decision metrics consistency are substantially improved:
Maximum entropy model can show good effect for limited training data, with Nae Bayesianmethod phase
Than its advantages are that do not have attribute conditions independence assumption, therefore select maximum entropy model, and failure menace level determines accuracy
It is high.
Based on maximum entropy model, increase by the way that the weight of software fault attribute is arranged, and to the fault data after participle
The prefix of fault attribute improves the training effect of maximum entropy model so that the input data of maximum entropy model is more reasonable.
In order to avoid single model be easy to cause the risk of over-fitting under limited training data, we use integrated study
Thought, the different fault attribute subset of option, the different model of training combines different moulds using effective combined strategy
Type prediction result, to obtain final result.
The cost of software fault menace level decision error is analyzed, for the prediction effect of effective assessment models, is used
Cost-sensitive error rate, so that the assessment of model is more in line with the actual conditions of application.
The above described is only a preferred embodiment of the present invention, be not intended to limit the present invention in any form, this
Field technical staff makes a little simple modification, equivalent variations or modification using the technology contents of the disclosure above, all falls within this hair
In bright protection scope.
Claims (10)
1. a kind of software fault menace level decision model training method characterized by comprising
According to the fault data of software fault, the multiple and different breakdown judge models based on maximum entropy theory of training；
According to the fault message of software fault, the multiple and different breakdown judge models based on maximum entropy theory of training, comprising:
Obtain the fault attribute of various software faults；
The fault attribute is converted into fault feature vector；
Utilize the fault feature vector, the multiple breakdown judge models based on maximum entropy theory of training.
2. software fault menace level decision model training method according to claim 1, which is characterized in that get
Different types of fault attribute is endowed different attribute weight.
3. software fault menace level decision model training method according to claim 2, which is characterized in that abstract type
Fault attribute be endowed attribute weight 3, the fault attribute of discovery time type is endowed attribute weight 1, software version type
Fault attribute be endowed attribute weight 1, the fault attribute for describing type is endowed attribute weight 2, the failure category of subject categories
Property is endowed attribute weight 2, and the fault attribute of annotation category is endowed attribute weight 1.
4. software fault menace level decision model training method according to claim 1, which is characterized in that in order to distinguish
Occur identical word in different types of attribute, fault attribute name will be added before each word as prefix.
5. software fault menace level decision model training method according to claim 1, which is characterized in that based on solely heat
Coding, is converted to fault feature vector for the fault attribute, comprising:
Judge whether to need to be segmented and gone stop words to the fault attribute got；
One-hot coding is carried out to the word sequence got by participle.
6. a kind of software fault menace level determination method characterized by comprising
Failure is predicted in real time using multiple breakdown judge models based on maximum entropy theory；
The real-time prediction result of multiple breakdown judge models is weighted and averaged, the fault level to software fault is obtained
Judgement.
7. software fault menace level determination method according to claim 6, which is characterized in that sentence to multiple failures
The real-time prediction result of disconnected model is weighted and averaged, and obtains the fault level judgement to software fault, comprising:
It is weighted and averaged according to following calculation formula, to obtain the fault level judgement to software fault:
Wherein, wiIt is the weight of model i, hi(x) be model prediction result.
8. software fault menace level determination method according to claim 7, which is characterized in that the weight of model i passes through
Following formula learns to obtain:
Wherein, accuracy rate of the model i in test data set j is cij。
9. a kind of software fault menace level decision model training device characterized by comprising
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Now according to claim 1 to software fault menace level decision model training method described in 5 any one.
10. a kind of software fault menace level decision maker characterized by comprising
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Software fault menace level determination method described in existing claim 6 to 8 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910805751.3A CN110502445B (en) | 2019-08-29 | 2019-08-29 | Software fault severity level judging method and device and model training method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910805751.3A CN110502445B (en) | 2019-08-29 | 2019-08-29 | Software fault severity level judging method and device and model training method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110502445A true CN110502445A (en) | 2019-11-26 |
| CN110502445B CN110502445B (en) | 2023-08-08 |
Family
ID=68590349
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910805751.3A Active CN110502445B (en) | 2019-08-29 | 2019-08-29 | Software fault severity level judging method and device and model training method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110502445B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112882887A (en) * | 2021-01-12 | 2021-06-01 | 昆明理工大学 | Dynamic establishment method for service fault model in cloud computing environment |
| CN113159090A (en) * | 2021-01-20 | 2021-07-23 | 渭南双盈未来科技有限公司 | Model self-adaptive information processing system and processing method thereof |
| CN114756460A (en) * | 2022-04-14 | 2022-07-15 | 中国电子科技集团公司第十五研究所 | Software failure mode judging method and system based on semantics |
| CN116125298A (en) * | 2022-11-28 | 2023-05-16 | 伏瓦科技(苏州)有限公司 | Battery fault detection method and device |
| CN117370066A (en) * | 2023-12-08 | 2024-01-09 | 杭州沃趣科技股份有限公司 | Method, device, equipment and storage medium for recovering server cluster |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103257921A (en) * | 2013-04-16 | 2013-08-21 | 西安电子科技大学 | Improved random forest algorithm based system and method for software fault prediction |
| CN106126512A (en) * | 2016-04-13 | 2016-11-16 | 北京天融信网络安全技术有限公司 | The Web page classification method of a kind of integrated study and device |
| CN106203534A (en) * | 2016-07-26 | 2016-12-07 | 南京航空航天大学 | A kind of cost-sensitive Software Defects Predict Methods based on Boosting |
| WO2019057363A1 (en) * | 2017-09-21 | 2019-03-28 | Thomson Licensing | Apparatus and method for rare failure prediction |
-
2019
- 2019-08-29 CN CN201910805751.3A patent/CN110502445B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103257921A (en) * | 2013-04-16 | 2013-08-21 | 西安电子科技大学 | Improved random forest algorithm based system and method for software fault prediction |
| CN106126512A (en) * | 2016-04-13 | 2016-11-16 | 北京天融信网络安全技术有限公司 | The Web page classification method of a kind of integrated study and device |
| CN106203534A (en) * | 2016-07-26 | 2016-12-07 | 南京航空航天大学 | A kind of cost-sensitive Software Defects Predict Methods based on Boosting |
| WO2019057363A1 (en) * | 2017-09-21 | 2019-03-28 | Thomson Licensing | Apparatus and method for rare failure prediction |
Non-Patent Citations (4)
| Title |
|---|
| 严蕾: "基于改进的随机森林的软件故障预测模型研究" * |
| 崔秀芝 * |
| 崔秀芝: "基于信息熵的多模型建模方法研究" * |
| 罗云锋等: "有限标注数据下的软件故障倾向预测方法", 《武汉理工大学学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112882887A (en) * | 2021-01-12 | 2021-06-01 | 昆明理工大学 | Dynamic establishment method for service fault model in cloud computing environment |
| CN112882887B (en) * | 2021-01-12 | 2022-08-09 | 昆明理工大学 | Dynamic establishment method for service fault model in cloud computing environment |
| CN113159090A (en) * | 2021-01-20 | 2021-07-23 | 渭南双盈未来科技有限公司 | Model self-adaptive information processing system and processing method thereof |
| CN114756460A (en) * | 2022-04-14 | 2022-07-15 | 中国电子科技集团公司第十五研究所 | Software failure mode judging method and system based on semantics |
| CN114756460B (en) * | 2022-04-14 | 2024-04-09 | 中国电子科技集团公司第十五研究所 | Semantic-based software fault mode judging method and system |
| CN116125298A (en) * | 2022-11-28 | 2023-05-16 | 伏瓦科技(苏州)有限公司 | Battery fault detection method and device |
| CN117370066A (en) * | 2023-12-08 | 2024-01-09 | 杭州沃趣科技股份有限公司 | Method, device, equipment and storage medium for recovering server cluster |
| CN117370066B (en) * | 2023-12-08 | 2024-03-15 | 杭州沃趣科技股份有限公司 | Method, device, equipment and storage medium for recovering server cluster |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110502445B (en) | 2023-08-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110502445A (en) | Software fault menace level determination method and device, model training method and device | |
| CN108564104A (en) | Product defects detection method, device, system, server and storage medium | |
| TW202014940A (en) | Training sample obtaining method, account prediction method, and corresponding devices | |
| CN109102157A (en) | A kind of bank's work order worksheet processing method and system based on deep learning | |
| CN112859822A (en) | Equipment health analysis and fault diagnosis method and system based on artificial intelligence | |
| US20130080834A1 (en) | Computer product, test support method, and test support apparatus | |
| CN104778186A (en) | Method and system for hanging commodity object to standard product unit (SPU) | |
| CN108255707A (en) | Development roles creation method, device, equipment and the storage medium of test case | |
| CN111797022A (en) | Test case generation method and device for order splitting service, electronic equipment and medium | |
| CN108255706A (en) | Edit methods, device, terminal device and the storage medium of automatic test script | |
| CN107833311A (en) | A kind of fault detection method and platform of shared bicycle | |
| WO2020146041A1 (en) | Change control in model-based system engineering | |
| EP3958130A1 (en) | Intelligent software testing | |
| CN107436916A (en) | The method and device of intelligent prompt answer | |
| Unterkalmsteiner et al. | Large-scale information retrieval in software engineering-an experience report from industrial application | |
| CN110516048A (en) | The extracting method, equipment and storage medium of list data in pdf document | |
| CN109376079A (en) | The test method and server that interface calls | |
| CN103514151A (en) | Dependency grammar analysis method and device and auxiliary classifier training method | |
| CN110990285A (en) | UI automation test method and device | |
| CN106997322A (en) | Method and apparatus for automatic test | |
| CN110716875A (en) | Concurrency test method based on feedback mechanism in domestic office environment | |
| CN109740947A (en) | Expert's method for digging, system, storage medium and electric terminal based on patent data | |
| CN104657248A (en) | Java thread stack analysis method and system | |
| CN106598843B (en) | A kind of software log behavior automatic identifying method based on program analysis | |
| CN106356994A (en) | Grid stability judging method based on grid PMU big data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |