CN108829607A - A kind of Software Defects Predict Methods based on convolutional neural networks - Google Patents
A kind of Software Defects Predict Methods based on convolutional neural networks Download PDFInfo
- Publication number
- CN108829607A CN108829607A CN201810743379.3A CN201810743379A CN108829607A CN 108829607 A CN108829607 A CN 108829607A CN 201810743379 A CN201810743379 A CN 201810743379A CN 108829607 A CN108829607 A CN 108829607A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- vector
- feature
- software
- 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
- G06F11/3604—Software analysis for verifying properties of programs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Abstract
A kind of Software Defects Predict Methods based on convolutional neural networks disclosed by the invention, include the following steps：Each file source code in analysis software project forms AST Token vector set；Mapping is established between integer and Token, and AST Token vector is converted into numerical value vector；Classify the processing of unbalanced problem using SMOTE technology logarithm vector collective data；Convolutional neural networks are constructed on the basis of numerical value vector set, extract the feature vector for capableing of expression code semanteme；The feature that convolutional neural networks learn is merged with traditional-handwork static nature；To there is the data acquisition system for merging feature to be input in support vector machine classifier, train software defect prediction model.The present invention may be directly applied in the failure prediction task of actual software, can capture the semantic feature of source code, solve the problems, such as that missing is to analysis of semantic characteristics in conventional method, and then improve the accuracy of bug prediction model.
Description
Technical field
It is analyzed the present invention relates to the software in soft project and failure prediction field, in particular to one kind is based on convolutional Neural
The Software Defects Predict Methods of network.
Background technique
Potential and unknown defect will seriously affect the quality of software, therefore software analysis and failure prediction skill in software
Art plays an important role in Software Quality Assurance task.If software defect can early be found, this will be helpful to soft
Part team understands the quality state of current project, and then reasonable distribution test resource.However to code units all in project
Examine it is unpractical manually, therefore, more and more soft project researchs and practitioner start to pay high attention to based on machine
The software defect Predicting Technique of device study, and attempting to come using a variety of machine learning methods may defective mould in inspection software
Block and file.
Software Defects Predict Methods based on machine learning extract code characteristic from the source file of software project first,
It mainly includes the Halstead feature based on operator and operand, the McCabe feature based on dependence, based on towards right
CK feature as program and the Change feature based on code change history etc..What these features mainly utilized is software analytics
The manual feature that persons refine.Further, Software Defects Predict Methods are using machine learning algorithm to the defective data of history
It carries out supervised learning and obtains disaggregated model, i.e., on the basis of carrying out manual feature extraction to historical data, with such as support
The study of the bases classifier such as vector machine, random forest, logistic regression progress bug prediction model.Prediction result is by helper applications matter
It may include the code region of defect that amount, which ensures that team finds,.
However, the manual feature that traditional Software Defects Predict Methods based on machine learning are utilized can not be captured and be utilized
Semantic feature abundant in software code, therefore estimated performance is often not satisfactory.Convolutional neural networks are raw as automation feature
At one of powerful technology, can effectively capture nonlinear characteristic complicated in software code, it is pre- to can solve software defect
Manual feature lacks the problem of semantic feature in survey method.
But convolutional neural networks how are realized in the application of software defect detection field, this is to those skilled in the art
For, it is a problem.
Summary of the invention
The purpose of the present invention is to overcome the shortcomings of the existing technology and deficiency, provides a kind of based on the soft of convolutional neural networks
Part failure prediction method, this method combination neural network automate the technology that feature generates, solve existing software defect prediction side
Traditional manual feature lacks the problem of semantic feature in method.This method operation is simple, only need to be in the pre- flow measurement of software defect of specification
It is that can provide a set of prediction result to each file defect of software project tendency that a software source codes are submitted on the basis of journey, is soft
Part team provides reference frame in project quality assessment and test resource distribution task.
The purpose of the present invention is realized by the following technical solution：
A kind of Software Defects Predict Methods based on convolutional neural networks, include the following steps：
1) each file source code in analysis software project obtains the AST (Abstract Syntax Trees) of each file
Token vector forms AST Token vector set；
2) mapping is established between integer and Token, the vector in AST Token vector set that step 1) is obtained turns
Change numerical value vector required for convolutional neural networks input into；
3) using a small number of classification data oversampling technologies (SMOTE), numerical value vector collective data that step 2) is obtained into
The processing of the unbalanced problem of row classification, final resulting result are balanced data set；
4) on the basis of passing through the numerical value vector set for unbalanced issue handling of classifying in step 3), convolutional Neural is constructed
Network extracts the feature vector for capableing of expression code semanteme；
5) feature vector learnt convolutional neural networks in step 4) and the Halstead based on operator and operand
Feature, the McCabe feature based on dependence, the CK feature based on object-oriented program merge, and form more complete static generation
Code feature vector, and then obtain the data acquisition system based on this feature vector；
6) data acquisition system obtained in step 5) is input in support vector machines (SVM) classifier, is trained based on volume
The software defect prediction model of product neural network.
The step 3) is specially：
Training data is pre-processed by a small number of classification data oversampling technologies (SMOTE)；SMOTE belongs to a kind of base
In the unbalanced data processing technique of the classification of over-sampling.It synthesizes new a few sample before model training, is learning tasks
The category distribution of balance is provided.
The oversampling technique, specific process are as follows：
A, for each of a small number of classifications sample, its k neighbour is found；
B, it is randomly choosed from each minority class very this k neighbours according to sample imbalance ratio；Choose neighbour's number
It is calculated according to classification disequilibrium rate；
Given M is as minority class very this quantity, and using N as the quantity of majority classification samples, oversampling technique
Strategy is to calculate uneven ratio IR=M/N；The neighbour number rk=1/IR-1 randomly selected, can not such as eliminate, then take upwards
It is whole；
C, the minority class of synthesis is very originally by very originally carrying out interpolation between the neighbour of selection in minority class.
The step 4) is specially：
The equalization data collection obtained by using step 3) trains a set of convolutional neural networks, i.e. study convolutional Neural net
Weight and deviation in network；
Specifically, the convolutional neural networks are by an input layer, two convolutional layers, two average pond layers and one
The hidden layer being fully connected forms, and all uses ReLU function as activation primitive in seven layers mentioned herein；The convolutional Neural
Network realized using Keras tool, constructs that extract after the convolutional neural networks with two layers of convolutional layer being capable of static code
Semantic expression characteristic.
Compared with prior art, the present invention having the following advantages that and beneficial effect：
Present invention utilizes the generations that convolutional neural networks carry out software source codes automation feature, can effectively capture source
Complicated nonlinear characteristic in code solves the problems, such as that missing is to analysis of semantic characteristics in Software Defects Predict Methods, and then improves
The accuracy of bug prediction model.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the Software Defects Predict Methods based on convolutional neural networks of the present invention.
Fig. 2 is the schematic diagram of AST Token vector analysis example.
Fig. 3 is the flow diagram that convolutional neural networks obtain semantic meaning representation feature.
Fig. 4 is the detailed maps of the pre- flow gauge of software defect based on convolutional neural networks.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
Such as Fig. 1,2,3,4, a kind of Software Defects Predict Methods based on convolutional neural networks include the following steps：
Each file source code in step 1) analysis software project obtains the Abstract Syntax Trees of each file
(AST) Token vector forms AST Token vector set.It is implemented as follows：The present invention selects in the AST of source code file
Parsing granularity of the node as vector.Carry out the source code of analysis software file using the open source library Java of an entitled JDT-core
Into AST Token vector.We mainly select the node of three types as label on AST：1) declaration node (contains method
Statement, type declarations etc.), value is extracted as the value of Token, 2) node of method call and the creation of class example, their quilts
It is recorded as their method name or class name, is put into Token and 3) includes If sentence, While sentence, For sentence, Throw
The controls stream node such as sentence is simply recorded the node type as them.The present invention selectes main AST node in table 1
In list.In this way, source file each in software project is converted to a Token vector by us.Generation as shown in Figure 2
Shown in code example, code segment can be parsed into AST Token vector：[For, myMethod, If, Continue, print].
The AST Token vector that step 1) obtains is converted into numerical value required for the input of convolutional neural networks by step 2)
Vector.Because the input of convolutional neural networks needs to be numerical value vector set, which needs will be obtained in step 1)
AST Token vector is converted into numerical value vector.In order to solve this problem, we establish first between integer and Token and reflect
It penetrates, and then Token vector is converted into integer vectors.Each Token is associated with unique integer identifiers, integer identifiers
Range from 1 to Token node type sum.In addition, convolutional neural networks need the vector length having the same inputted, but
Be the last integer vectors of different files length it is often different.Therefore the tail portion that we are attached to each integer vectors for 0, makes
Its length is consistent with longest integer vectors.In addition, occurring Token section two or more times in our processing software projects
Point, and primary Token node will only occur and be expressed as 0, avoid file from leading to one because calling the sentence seldom occurred with this
The integer vectors having little significance a bit.Listed in table 2 three AST Token vectors set and its be converted into integer vectors
Example, wherein Continue sentence only occurred once in set, therefore we do not consider it, in first integer vectors
Marked it for 0, Article 3 AST Token vector is maximum length sequence, length 7, therefore first and Article 2 integer to
Measure 0 polishing length to 7.
Table 1：The selected main AST node of the present invention
Table 2：AST Token vector is converted into numerical value vector
The numerical value that step 3) is obtained using to the over-samplings of a small number of categorical datas (using SMOTE method), to step 2) to
Duration set data classify the processing of unbalanced problem.In fact, the ratio of defects of some software projects is very low or very high, institute
With in the software defect forecasting problem based on machine learning, classification is unbalanced to be generally existing and has the problem of challenge.Such as
We carry out the training of disaggregated model to unbalanced data set of highly classifying to fruit, and class imbalance problem may come to learning tape
Difficulty, because training the classifier come tends to the most classification of selection, and weaker to the other classification capacity of minority class.For
Alleviation this problem, unbalanced learning art of classifying are widely used.In the present invention, we are before prediction model study
Model training data are pre-processed using a small number of classification data oversampling technologies (SMOTE).SMOTE herein is a kind of mistake
Sampling techniques is spent, the other new samples of minority class can be synthetically created, more balanced classification distribution is provided for learning tasks.
On the basis of the numerical value vector set of processing of the step 4) in step 3) by unbalanced problem of classifying, building volume
Product neural network, extracts the expression characteristic for capableing of static code semanteme.The equalization data that we are obtained by using step 3)
Collect to train a set of convolutional neural networks, i.e. weight and deviation in study convolutional neural networks.Specifically, our convolution
Neural network is made of an input layer, two convolutional layers, two average pond layers and a hidden layer being fully connected, here
All use ReLU function as activation primitive in seven layers mentioned.Fig. 3 is that convolutional neural networks obtain semantic feature in the step
The schematic diagram of process.That the realization of convolutional neural networks utilizes in the present invention is Keras tool (http://keras.io).It is logical
Convolutional neural networks can rapidly be constructed by crossing Keras, and obtain the feature vector that convolutional neural networks are extract.
The feature vector that step 5) learns convolutional neural networks in step 4) with based on operator and operand
Halstead feature, the McCabe feature based on dependence, (table 3 lists allusion quotation to the CK feature merging based on object-oriented program
Halstead, McCabe and CK feature of type), more complete static code feature vector is formed, and obtain based on this feature
The data acquisition system of vector.Until step 4), we only consider the semantic-based feature extracted by convolutional neural networks.
However, in traditional failure prediction method, other features such as complexity measure and object-oriented program feature also contain can
For predicting the static code characteristic information of defect.These static natures can be obtained by code analysis.In order to utilize this
A little information, the feature vector (feature vector representated by the hidden layer being fully connected) and pass that we learn convolutional neural networks
The static nature vector of system connects.This connection can be realized by the union operator in Keras.Include in Fig. 4
The step schematic diagram that the feature that convolutional neural networks are extract merges with traditional-handwork feature.Merge completing feature vector
We obtain the data acquisition system based on this feature vector afterwards.The step also carries out Z- to data acquisition system after the completion of aforesaid operations
Score standardization.
Table 3：Representational Halstead, McCabe and CK feature
| Halstead&McCabe feature | CK feature |
| Operator quantity and number of operations | The method quantity of class |
| Operation object quantity and by number of operations | Depth of the class in inheritance tree |
| Essential complexity | Child nodes number of the class in inheritance tree |
| Annular complexity | There are the quantity of other classes of coupled relation with class |
| Design complexities | The externalist methodology quantity that can be called in class |
| Lines of code | The side operator of one or more attributes is accessed in class |
6) data acquisition system obtained in step 5) is input in support vector machines (SVM) classifier, is trained based on volume
The software defect prediction model of product neural network.The present invention utilizes open source library Libsvm (https://
Www.csie.ntu.edu.tw/~cjlin/libsvm/) realize SVM classifier training.The last one is as supporting vector
The convolutional neural networks unit output layer of machine classifier uses Sigmoid as activation primitive.Fig. 4 is to realize based on convolution mind
The detailed maps of the pre- flow gauge of software defect through network.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (3)
1. a kind of Software Defects Predict Methods based on convolutional neural networks, which is characterized in that include the following steps：
1) each file source code in analysis software project, obtains the AST Token vector of each file, formed AST Token to
Duration set；
2) mapping is established between integer and Token, and the vector in AST Token vector set that step 1) obtains is converted into
Numerical value vector required for convolutional neural networks input；
3) using a small number of classification data oversampling technologies, the numerical value vector collective data that step 2) obtains classify unbalanced
The processing of problem, final resulting result are balanced data set；
4) on the basis of passing through the numerical value vector set for unbalanced issue handling of classifying in step 3), convolutional Neural net is constructed
Network extracts the feature vector for capableing of expression code semanteme；
5) feature vector and the Halstead feature based on operator and operand learnt convolutional neural networks in step 4),
McCabe feature based on dependence, the CK feature based on object-oriented program merge, and it is special to form more complete static code
Vector is levied, and then obtains the data acquisition system based on this feature vector；
6) data acquisition system obtained in step 5) is input in support vector machine classifier, is trained based on convolutional neural networks
Software defect prediction model.
2. according to claim 1 based on the Software Defects Predict Methods of convolutional neural networks, which is characterized in that the step
3) it is specially：
Training data is pre-processed by a small number of classification data oversampling technologies；
The oversampling technique, specific process are as follows：
A, for each of a small number of classifications sample, its k neighbour is found；
B, it is randomly choosed from each minority class very this k neighbours according to sample imbalance ratio；Choose neighbour's number according to
Classification disequilibrium rate calculates；
Quantity of the given M as minority class very originally, and using N as the quantity of most classification samples, the strategy of oversampling technique
It is to calculate uneven ratio IR=M/N；The neighbour number rk=1/IR-1 randomly selected, can not such as eliminate, then round up；
C, the minority class of synthesis is very originally by very originally carrying out interpolation between the neighbour of selection in minority class.
3. according to claim 1 based on the Software Defects Predict Methods of convolutional neural networks, which is characterized in that the step
4) it is specially：
The equalization data collection obtained by using step 3) trains a set of convolutional neural networks, i.e., in study convolutional neural networks
Weight and deviation；
Specifically, the convolutional neural networks are complete by an input layer, two convolutional layers, two average pond layers and one
The hidden layer of connection forms, and all uses ReLU function as activation primitive in seven layers mentioned herein；The convolutional neural networks
Realized using Keras tool, construct extract after the convolutional neural networks with two layers of convolutional layer can static code it is semantic
Expression characteristic.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810743379.3A CN108829607B (en) | 2018-07-09 | 2018-07-09 | Software defect prediction method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810743379.3A CN108829607B (en) | 2018-07-09 | 2018-07-09 | Software defect prediction method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108829607A true CN108829607A (en) | 2018-11-16 |
| CN108829607B CN108829607B (en) | 2021-08-10 |
Family
ID=64136032
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810743379.3A Active CN108829607B (en) | 2018-07-09 | 2018-07-09 | Software defect prediction method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108829607B (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109376092A (en) * | 2018-11-26 | 2019-02-22 | 扬州大学 | A kind of software defect reason automatic analysis method of facing defects patch code |
| CN109710512A (en) * | 2018-12-06 | 2019-05-03 | 南京邮电大学 | Neural network software failure prediction method based on geodesic curve stream core |
| CN110109835A (en) * | 2019-05-05 | 2019-08-09 | 重庆大学 | A kind of software defect positioning method based on deep neural network |
| CN110134108A (en) * | 2019-05-14 | 2019-08-16 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | A kind of aacode defect test method and device |
| CN110489348A (en) * | 2019-08-23 | 2019-11-22 | 山东浪潮人工智能研究院有限公司 | A kind of software function defect method for digging based on transfer learning |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN110825615A (en) * | 2019-09-23 | 2020-02-21 | 中国科学院信息工程研究所 | Software defect prediction method and system based on network embedding |
| CN111522743A (en) * | 2020-04-17 | 2020-08-11 | 北京理工大学 | Software defect prediction method based on gradient lifting tree support vector machine |
| CN111767216A (en) * | 2020-06-23 | 2020-10-13 | 江苏工程职业技术学院 | Cross-version depth defect prediction method capable of relieving class overlap problem |
| CN112070716A (en) * | 2020-08-03 | 2020-12-11 | 西安理工大学 | Printing defect intelligent identification method based on deep learning |
| CN112286807A (en) * | 2020-10-28 | 2021-01-29 | 北京航空航天大学 | Software defect positioning system based on source code file dependency relationship |
| CN112328469A (en) * | 2020-10-22 | 2021-02-05 | 南京航空航天大学 | Function level defect positioning method based on embedding technology |
| CN112631898A (en) * | 2020-12-09 | 2021-04-09 | 南京理工大学 | Software defect prediction method based on CNN-SVM |
| CN112905468A (en) * | 2021-02-20 | 2021-06-04 | 华南理工大学 | Ensemble learning-based software defect prediction method, storage medium and computing device |
| CN113608747A (en) * | 2021-08-18 | 2021-11-05 | 南京航空航天大学 | Software defect prediction method based on graph convolution neural network and terminal |
| CN113722218A (en) * | 2021-08-23 | 2021-11-30 | 南京审计大学 | Software defect prediction model construction method based on compiler intermediate representation |
| CN114169500A (en) * | 2021-11-30 | 2022-03-11 | 电子科技大学 | Neural network model processing method based on small-scale electromagnetic data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105930277A (en) * | 2016-07-11 | 2016-09-07 | 南京大学 | Defect source code locating method based on defect report analysis |
| US20170212829A1 (en) * | 2016-01-21 | 2017-07-27 | American Software Safety Reliability Company | Deep Learning Source Code Analyzer and Repairer |
| CN107103235A (en) * | 2017-02-27 | 2017-08-29 | 广东工业大学 | A kind of Android malware detection method based on convolutional neural networks |
| CN107885999A (en) * | 2017-11-08 | 2018-04-06 | 华中科技大学 | A kind of leak detection method and system based on deep learning |
-
2018
- 2018-07-09 CN CN201810743379.3A patent/CN108829607B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170212829A1 (en) * | 2016-01-21 | 2017-07-27 | American Software Safety Reliability Company | Deep Learning Source Code Analyzer and Repairer |
| CN105930277A (en) * | 2016-07-11 | 2016-09-07 | 南京大学 | Defect source code locating method based on defect report analysis |
| CN107103235A (en) * | 2017-02-27 | 2017-08-29 | 广东工业大学 | A kind of Android malware detection method based on convolutional neural networks |
| CN107885999A (en) * | 2017-11-08 | 2018-04-06 | 华中科技大学 | A kind of leak detection method and system based on deep learning |
Non-Patent Citations (3)
| Title |
|---|
| JIAN LI等: "Software Defect Prediction via Convolutional Neural Network", 《 2017 IEEE INTERNATIONAL CONFERENCE ON SOFTWARE QUALITY, RELIABILITY AND SECURITY (QRS)》 * |
| SONG WANG等: "Automatically Learning Semantic Features for Defect Prediction", 《2016 IEEE/ACM 38TH INTERNATIONAL CONFERENCE ON SOFTWARE ENGINEERING (ICSE)》 * |
| 李克文等: "基于改进ACO优化BPNN的软件缺陷预测模型", 《计算机工程与设计》 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109376092A (en) * | 2018-11-26 | 2019-02-22 | 扬州大学 | A kind of software defect reason automatic analysis method of facing defects patch code |
| CN109710512A (en) * | 2018-12-06 | 2019-05-03 | 南京邮电大学 | Neural network software failure prediction method based on geodesic curve stream core |
| CN110109835A (en) * | 2019-05-05 | 2019-08-09 | 重庆大学 | A kind of software defect positioning method based on deep neural network |
| CN110109835B (en) * | 2019-05-05 | 2021-03-30 | 重庆大学 | Software defect positioning method based on deep neural network |
| CN110134108A (en) * | 2019-05-14 | 2019-08-16 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | A kind of aacode defect test method and device |
| CN110134108B (en) * | 2019-05-14 | 2021-10-22 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | Code defect testing method and device |
| CN110489348A (en) * | 2019-08-23 | 2019-11-22 | 山东浪潮人工智能研究院有限公司 | A kind of software function defect method for digging based on transfer learning |
| CN110825615A (en) * | 2019-09-23 | 2020-02-21 | 中国科学院信息工程研究所 | Software defect prediction method and system based on network embedding |
| CN110597735B (en) * | 2019-09-25 | 2021-03-05 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN110597735A (en) * | 2019-09-25 | 2019-12-20 | 北京航空航天大学 | Software defect prediction method for open-source software defect feature deep learning |
| CN111522743A (en) * | 2020-04-17 | 2020-08-11 | 北京理工大学 | Software defect prediction method based on gradient lifting tree support vector machine |
| CN111522743B (en) * | 2020-04-17 | 2021-10-22 | 北京理工大学 | Software defect prediction method based on gradient lifting tree support vector machine |
| CN111767216A (en) * | 2020-06-23 | 2020-10-13 | 江苏工程职业技术学院 | Cross-version depth defect prediction method capable of relieving class overlap problem |
| CN112070716A (en) * | 2020-08-03 | 2020-12-11 | 西安理工大学 | Printing defect intelligent identification method based on deep learning |
| CN112328469A (en) * | 2020-10-22 | 2021-02-05 | 南京航空航天大学 | Function level defect positioning method based on embedding technology |
| CN112286807B (en) * | 2020-10-28 | 2022-01-28 | 北京航空航天大学 | Software defect positioning system based on source code file dependency relationship |
| CN112286807A (en) * | 2020-10-28 | 2021-01-29 | 北京航空航天大学 | Software defect positioning system based on source code file dependency relationship |
| CN112631898A (en) * | 2020-12-09 | 2021-04-09 | 南京理工大学 | Software defect prediction method based on CNN-SVM |
| CN112905468A (en) * | 2021-02-20 | 2021-06-04 | 华南理工大学 | Ensemble learning-based software defect prediction method, storage medium and computing device |
| CN113608747A (en) * | 2021-08-18 | 2021-11-05 | 南京航空航天大学 | Software defect prediction method based on graph convolution neural network and terminal |
| CN113608747B (en) * | 2021-08-18 | 2024-04-02 | 南京航空航天大学 | Software defect prediction method and terminal based on graph convolution neural network |
| CN113722218A (en) * | 2021-08-23 | 2021-11-30 | 南京审计大学 | Software defect prediction model construction method based on compiler intermediate representation |
| CN114169500A (en) * | 2021-11-30 | 2022-03-11 | 电子科技大学 | Neural network model processing method based on small-scale electromagnetic data |
| CN114169500B (en) * | 2021-11-30 | 2023-04-18 | 电子科技大学 | Neural network model processing method based on small-scale electromagnetic data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108829607B (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108829607A (en) | A kind of Software Defects Predict Methods based on convolutional neural networks | |
| CN106951925A (en) | Data processing method, device, server and system | |
| CN110597735A (en) | Software defect prediction method for open-source software defect feature deep learning | |
| CN110119446A (en) | Interpretable artificial intelligence | |
| CN109408389A (en) | A kind of aacode defect detection method and device based on deep learning | |
| CN110399490A (en) | A kind of barrage file classification method, device, equipment and storage medium | |
| CN108563556A (en) | Software defect prediction optimization method based on differential evolution algorithm | |
| CN113159095A (en) | Model training method, image retrieval method and device | |
| CN109299258A (en) | A kind of public sentiment event detecting method, device and equipment | |
| CN110351301A (en) | A kind of double-deck progressive method for detecting abnormality of HTTP request | |
| Maiti et al. | Capturing, eliciting, predicting and prioritizing (CEPP) non-functional requirements metadata during the early stages of agile software development | |
| CN109872162A (en) | A kind of air control classifying identification method and system handling customer complaint information | |
| CN110610193A (en) | Method and device for processing labeled data | |
| CN103886030B (en) | Cost-sensitive decision-making tree based physical information fusion system data classification method | |
| WO2020228283A1 (en) | Feature extraction method and apparatus, and computer readable storage medium | |
| CN105469376A (en) | Method and device for determining picture similarity | |
| CN107463935A (en) | Application class methods and applications sorter | |
| CN113361258A (en) | Aspect-level emotion analysis method and system based on graph convolution network and attention selection | |
| CN110221975A (en) | Create the method and device of interface use-case automatic test script | |
| CN112270441A (en) | Method for establishing autism child rehabilitation effect prediction model and method and system for predicting autism child rehabilitation effect | |
| CN107958270A (en) | Classification recognition methods, device, electronic equipment and computer-readable recording medium | |
| Meilong et al. | An approach to semantic and structural features learning for software defect prediction | |
| CN109345262A (en) | It is a kind of intelligently to complain classification and processing system | |
| CN115358481A (en) | Early warning and identification method, system and device for enterprise ex-situ migration | |
| CN113608747A (en) | Software defect prediction method based on graph convolution neural network and terminal |
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 |