CN108647138A - A kind of Software Defects Predict Methods, device, storage medium and electronic equipment - Google Patents
A kind of Software Defects Predict Methods, device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN108647138A CN108647138A CN201810162377.5A CN201810162377A CN108647138A CN 108647138 A CN108647138 A CN 108647138A CN 201810162377 A CN201810162377 A CN 201810162377A CN 108647138 A CN108647138 A CN 108647138A
- Authority
- CN
- China
- Prior art keywords
- default
- data
- software
- data set
- prototype
- 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.)
- Pending
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
- G06F11/3608—Software analysis for verifying properties of programs using formal methods, e.g. model checking, abstract interpretation
Abstract
The invention discloses a kind of Software Defects Predict Methods, device, storage medium and electronic equipment, this method includes：The sample data for concentrating selection predetermined number in the first default initial data according to the first default selection rule, to obtain the first prototype data collection；The first data set of dissimilarity between the first default raw data set and the first prototype data collection is calculated according to the first pre-determined distance algorithm；Data in first data set are input to predetermined software bug prediction model, to obtain the software defect prediction result of the first default raw data set corresponding software, wherein shown predetermined software bug prediction model is according to the model for presetting dissimilarity structure.By that with the present invention, can determine that software whether there is defect according to dissimilarity, fundamentally promote estimated performance, accuracy rate is higher, and better user experience solves problem of the prior art.
Description
Technical field
The present invention relates to data processing field, more particularly to a kind of Software Defects Predict Methods, device, storage medium and
Electronic equipment.
Background technology
Defective sample size is therefore, soft often than flawless sample size much less in software defect data set
Part failure prediction can be considered a unbalanced problem concerning study of class.During class unbalanced study study, different classes of mistake
Divide cost also each unequal, wherein the accidentally point cost of minority class (defective) accidentally divides cost far above most classes (zero defect),
To reduce accidentally point cost to the maximum extent, prediction algorithm more focuses on being promoted the predictablity rate of defective minority class sample.It is real
On border, traditional sorting algorithm is typically established at class distributing equilibrium and under the premise of accidentally point cost is equal, is missed with minimizing classification
Difference is final goal, therefore directly uses decision tree classification, neural network, Bayes's classification, support vector machines and k- arest neighbors
Traditional machine learning algorithms such as classification can not obtain preferable software defect estimated performance.
In recent years, the unbalanced problem concerning study of class receives the extensive concern of academia, machine learning and Data Mining
Brainstrust proposes many effective solutions in terms of data Layer and algorithm layer two.
About data layer method, mainly by way of sampling or generating new samples, class distribution is made to restore balanced, it is such as random
Sub- sampling (RUS) and random oversampling (ROS).Duplicate sampling can be distributed with balanced class, but sub- sampling often ignore it is certain heavy
Sample is wanted, loss of learning is caused；Conversely, oversampling can introduce a large amount of copies, redundancy is generated, over-fitting is caused.
About algorithm layer method, lays particular emphasis on and improve existing sorting algorithm or study new sorting algorithm, preferably to solve
The unbalanced problem concerning study of class." One-Class Learning " method, this method only build disaggregated model in most classes, it is difficult to
Accurate Prediction minority class；Ensemble of learning methods builds multiple disaggregated models by duplicate sampling, iteration updates the power of training sample
Weight or the mode for combining multiple decision trees, obtain stable nicety of grading, such as Bagging, Boosting and Random Forest
Scheduling algorithm.In particular, when, there are when significant difference, assembled classification model is more more acurrate than basic classification model between disaggregated model, but
Its computationally intensive and complexity is higher；Cost-sensitive is analyzed, and accidentally cost is divided to be disobeyed as learning objective, such as MetaCost to minimize
Rely in sorting algorithm, and can be applied on any form of cost matrix, but how to determine cost matrix at present is still one
Problem.
Therefore, how the unbalanced learning method of existing class adjusts class distribution or innovatory algorithm if being laid particular emphasis on, can not be from basic
The upper estimated performance for promoting such problem, predictablity rate is relatively low, and user experience is poor.
Invention content
A kind of Software Defects Predict Methods of present invention offer, device, storage medium and electronic equipment, to solve existing skill
The following problem of art：How the existing unbalanced learning method of class adjusts class distribution or innovatory algorithm if being laid particular emphasis on, can not be from basic
The upper estimated performance for promoting such problem, predictablity rate is relatively low, and user experience is poor.
In order to solve the above technical problems, on the one hand, the present invention provides a kind of Software Defects Predict Methods, including：According to
One default selection rule concentrates the sample data of selection predetermined number in the first default initial data, to obtain the first prototype data
Collection；Not phase is calculated between the first default raw data set and the first prototype data collection according to the first pre-determined distance algorithm
Like the first data set of property；Data in first data set are input to predetermined software bug prediction model, to obtain
State the software defect prediction result of the first default raw data set corresponding software, wherein shown predetermined software bug prediction model
For according to the model for presetting dissimilarity structure.
Optionally, the sample number of selection predetermined number is concentrated in the first default initial data according to the first default selection rule
According to before obtaining the first prototype data collection, to further include：It is concentrated in the second default initial data according to the second default selection rule
The sample data for selecting predetermined number, to obtain the second prototype data collection；Described second is calculated according to the second pre-determined distance algorithm
Second data set of default dissimilarity between raw data set and the second prototype data collection；According to second data set
In data and default sorting algorithm build the predetermined software bug prediction model.
Optionally, the described first default selection rule or the second default selection rule include at least following one：With
Machine selection method, the linear programming method based on cluster.
Optionally, the first pre-determined distance algorithm or the second pre-determined distance algorithm include at least following one：Europe
Distance algorithm, manhatton distance algorithm, Minkowski Distance algorithm are obtained in several.
On the other hand, the present invention also provides a kind of software defect prediction meanss, including：First choice module, for according to
First default selection rule concentrates the sample data of selection predetermined number in the first default initial data, to obtain the first modal number
According to collection；
First computing module, for according to the first pre-determined distance algorithm calculate the first default raw data set with it is described
First data set of dissimilarity between first prototype data collection；Prediction module is used for the data in first data set
It is input to predetermined software bug prediction model, the software defect to obtain the described first default raw data set corresponding software is predicted
As a result, wherein shown predetermined software bug prediction model is according to the model for presetting dissimilarity structure.
Optionally, further include：Second selecting module is used for according to the second default selection rule in the second default initial data
The sample data for concentrating selection predetermined number, to obtain the second prototype data collection；Second computing module, for default according to second
Distance algorithm calculates the second data of dissimilarity between the described second default raw data set and the second prototype data collection
Collection；Module is built, for according to the data and the default sorting algorithm structure predetermined software defect in second data set
Prediction model.
Optionally, the described first default selection rule or the second default selection rule include at least following one：With
Machine selection method, the linear programming method based on cluster.
Optionally, the first pre-determined distance algorithm or the second pre-determined distance algorithm include at least following one：Europe
Distance algorithm, manhatton distance algorithm, Minkowski Distance algorithm are obtained in several.
On the other hand, the present invention also provides a kind of storage mediums, are stored with computer program, and the computer program is located
The step of reason device realizes above-mentioned Software Defects Predict Methods when executing.
On the other hand, memory, processor are included at least the present invention also provides a kind of electronic equipment, on the memory
It is stored with computer program, the processor realizes that above-mentioned software defect is pre- in the computer program on executing the memory
The step of survey method.
The present invention considers the discriminating power of attributive character in class imbalanced data sets, therefore, the original that selection is obtained
Type data set raw data set corresponding with the prototype data collection carries out dissimilarity calculating, and then determines that has a dissmilarity
Property data data set, and the data in the data set are input to the software defect prediction built previously according to dissimilarity
In model, it will be able to determine that software whether there is defect according to dissimilarity, fundamentally promote estimated performance, accuracy rate compared with
Height, better user experience solve the problems, such as the as follows of the prior art：How the existing unbalanced learning method of class adjusts if being laid particular emphasis on
Class be distributed or innovatory algorithm, can not fundamentally promote the estimated performance of such problem, predictablity rate is relatively low, user experience compared with
Difference.
Description of the drawings
Fig. 1 is the flow chart of Software Defects Predict Methods in first embodiment of the invention；
Fig. 2 is the structural schematic diagram of software defect prediction meanss in second embodiment of the invention；
Fig. 3 is the configuration diagram of software defect Prediction program in third embodiment of the invention.
Specific implementation mode
In order to solve the problems, such as the as follows of the prior art：How the existing unbalanced learning method of class adjusts class distribution if being laid particular emphasis on
Or innovatory algorithm, the estimated performance of such problem can not be fundamentally promoted, predictablity rate is relatively low, and user experience is poor；This
Invention provides a kind of Software Defects Predict Methods, device, storage medium and electronic equipment, below in conjunction with attached drawing and implementation
Example, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used to explain this hair
It is bright, do not limit the present invention.
First embodiment of the invention provides a kind of Software Defects Predict Methods, and the flow of this method is as shown in Figure 1, include
Step S101 to S103：
S101 concentrates the sample number of selection predetermined number according to the first default selection rule in the first default initial data
According to obtain the first prototype data collection；
S102 is calculated between the first default raw data set and the first prototype data collection not according to the first pre-determined distance algorithm
First data set of similitude；
Data in first data set are input to predetermined software bug prediction model by S103, to obtain the first default original
The software defect prediction result of beginning data set corresponding software, wherein shown predetermined software bug prediction model is that basis is default not
The model of similitude structure.
The embodiment of the present invention considers the discriminating power of attributive character in class imbalanced data sets, therefore, to selecting
The prototype data collection arrived raw data set corresponding with the prototype data collection carries out dissimilarity calculating, and then determining one has
The data set of dissimilarity data, and the data in the data set are input to the software built previously according to dissimilarity and are lacked
It falls into prediction model, it will be able to determine that software whether there is defect according to dissimilarity, fundamentally promote estimated performance, accurately
Rate is higher, better user experience, solves the problems, such as the as follows of the prior art：The existing unbalanced learning method of class lay particular emphasis on how
Class distribution or innovatory algorithm are adjusted, can not fundamentally promote the estimated performance of such problem, predictablity rate is relatively low, user's body
It tests poor.
The sample data for selecting predetermined number is being concentrated in the first default initial data according to the first default selection rule, with
Before obtaining the first prototype data collection, need first to build above-mentioned predetermined software bug prediction model, specific implementation process is also
It is calculated, is specifically included according to dissimilarity：It is concentrated in the second default initial data according to the second default selection rule
The sample data for selecting predetermined number, to obtain the second prototype data collection；It is default that second is calculated according to the second pre-determined distance algorithm
Second data set of dissimilarity between raw data set and the second prototype data collection；According to data in the second data set and pre-
If sorting algorithm builds predetermined software bug prediction model.By the above process, that is, predetermined software bug prediction model is completed
Structure.
It is predicted due to the building process of predetermined software bug prediction model and using predetermined software bug prediction model
Principle be identical, therefore, all there is similarity in the various rules and algorithms of above-mentioned use, for example, first default chooses
Rule or the second default selection rule include at least following one：Random selection method, the linear programming method based on cluster；The
One pre-determined distance algorithm or the second pre-determined distance algorithm include at least following one：Euclidean distance algorithm, manhatton distance
Algorithm, Minkowski Distance algorithm.
Second embodiment of the invention provides a kind of software defect prediction meanss, the structural representation of the device as shown in Fig. 2,
Including：
First choice module 10, for concentrating selection predetermined in the first default initial data according to the first default selection rule
The sample data of number, to obtain the first prototype data collection；First computing module 20, couples with first choice module 10, is used for
First of dissimilarity between the first default raw data set and the first prototype data collection is calculated according to the first pre-determined distance algorithm
Data set；Prediction module 30 is coupled with the first computing module 20, for the data in the first data set to be input to predetermined software
Bug prediction model, to obtain the software defect prediction result of the first default raw data set corresponding software, wherein shown default
Software defect prediction model is according to the model for presetting dissimilarity structure.
The embodiment of the present invention considers the discriminating power of attributive character in class imbalanced data sets, therefore, to selecting
The prototype data collection arrived raw data set corresponding with the prototype data collection carries out dissimilarity calculating, and then determining one has
The data set of dissimilarity data, and the data in the data set are input to the software built previously according to dissimilarity and are lacked
It falls into prediction model, it will be able to determine that software whether there is defect according to dissimilarity, fundamentally promote estimated performance, accurately
Rate is higher, better user experience, solves the problems, such as the as follows of the prior art：The existing unbalanced learning method of class lay particular emphasis on how
Class distribution or innovatory algorithm are adjusted, can not fundamentally promote the estimated performance of such problem, predictablity rate is relatively low, user's body
It tests poor.
Before above-mentioned module carries out software defect prediction, it is also necessary to above-mentioned predetermined software bug prediction model is first built,
Therefore, above-mentioned apparatus can also include the following modules coupled successively：Second selecting module, for according to the second default choosing
Rule is taken to concentrate the sample data of selection predetermined number in the second default initial data, to obtain the second prototype data collection；Second
Computing module, for being calculated between the second default raw data set and the second prototype data collection not according to the second pre-determined distance algorithm
Second data set of similitude；Build module, for according in the second data set data and default sorting algorithm build it is default
Software defect prediction model.
It is predicted due to the building process of predetermined software bug prediction model and using predetermined software bug prediction model
Principle be identical, therefore, all there is similarity in the various rules and algorithms of above-mentioned use, for example, first default chooses
Rule or the second default selection rule include at least following one：Random selection method, the linear programming method based on cluster；The
One pre-determined distance algorithm or the second pre-determined distance algorithm include at least following one：Euclidean distance algorithm, manhatton distance
Algorithm, Minkowski Distance algorithm.
In order to promote the discriminating power of software defect data set attribute feature, third embodiment of the invention provides a kind of electricity
There is on the storage medium of the electronic equipment equipment a set of software defect Prediction program, the program to use a kind of based on not
The representation of similitude substitutes primitive attribute feature with dissimilarity between sample, not only remains the original statistical information of data set,
Also can get in data set structural information, and then promote the estimated performance of software defect.
When using machine learning algorithm forecasting software defect, the foundation of prediction model is typically based on the soft of staticametric member
On part defective data collection, and the present invention is then that raw data set is mapped to dissimilarity space in advance, then in dissimilarity
Software defect prediction model is built in space.The embodiment of the present invention is mainly converted by prototype selection, dissimilarity and classification three
It is grouped as.Fig. 3 gives the overall framework of the embodiment of the present invention, mainly by structure and software based on dissimilarity prediction model
Two big link of failure prediction forms.
(1) structure based on dissimilarity prediction model.
The building process of software defect prediction algorithm based on dissimilarity mainly by prototype selection, dissimilarity conversion and
Build prediction model three parts composition.First, it is filtered out from initial data concentration using prototype selection method representative
Sample creates prototype collection as prototype；Then, the dissimilarity between raw data set and prototype collection sample is calculated, thus by it
It is mapped in corresponding dissimilarity space；Finally, using traditional classification algorithm software defect is built in dissimilarity space
Prediction model.
1, prototype selection.
Prototype selection, which is intended to concentrate from initial data, chooses representative sample as prototype, turns as dissimilarity
Reference when changing.In order to preferably choose prototype, scholars propose based on shared arest neighbors (Shared Nearest
Neighbors, SNN) Jarvis-Patrick clustering (JPC) algorithm, random selection (RandomC, RC), linear
Plan (LinPro), Attributions selection FeaSel), pattern search (ModeSeek), the linear programming (KCenters- based on cluster
) and the methods of editing compressed (EdiCon) LP.Wherein, random choice method (RC) randomly concentrates extraction to refer to from initial data
The sample of fixed number amount is a kind of most simple and effective prototype selection method as prototype.Pekalska et al. comparative analyses
Above-mentioned prototype method to the influence based on dissimilarity sorting technique performance, the experimental results showed that：The total body surfaces of RC and KCenters
Now preferably, but RC is more convenient.
In order to ensure the estimated performance of the present invention, select RC as prototype selection method, from priginal soft defective data collection
It is middle to extract representative sample, create prototype collection.Assuming that D represents a software defect data set, belongs to two classification and ask
Topic, i.e. C={ c1, c2, DtFor training set, D1And D2Respectively represent defective and zero defect class training set.R sample is extracted from D
This is as prototype set P, using randomly selected method respectively from D1And D2In randomly select r1And r2A sample makes prototype collection P
={ r1,r2}。
2, dissimilarity is converted.
Dissimilarity conversion is intended to raw data set being mapped to dissimilarity space.
Assuming that D={ x1,x2,…,xnRepresentative sample quantity be n software defect data set.Wherein, xi={ ai1,
ai2,…,aim,ciRepresent i-th of sample in data set D；xiIt is made of m independent attribute and a generic attribute.P={ p1,
p2,…,prIndicate the prototype collection being made of r representative samples, wherein pi={ a 'i1,a′i2,…,a′im,c′iGeneration
I-th of prototype of table.
With reference to prototype collection P, by calculating the dissimilarity in original software defect data set D and prototype collection P between sample,
Data set D is mapped to corresponding dissimilarity space, obtains the data set based on dissimilarity It is made of n sample.Each sampleBelonged to by r+1
Property composition, whereinRepresentative sample xiWith prototype pjBetween dissimilarity.
When dissimilarity between assessing intensive, continuous type sample, the dissimilarity between the distance sample based on measurement is logical
Common distance measures to describe, and distance is bigger, more dissimilar；Conversely, then more similar.Currently, most common distance metric has Europe
A few Reed distances, manhatton distance and Minkowski Distance.Wherein, Minkowski Distance is Euclidean distance and Man Ha
The popularization for distance of pausing, computational methods are shown in formula：
In formula, l is real number, l >=1.
As l=1, manhatton distance, i.e. L1Norm；
As l=2, Euclidean distance, i.e. L2Norm is usually used in measuring in intensive, continuous data set between sample
Dissimilarity；
As l=∞, supremum distance, also known as L∞Norm and Chebyshev's distance measure the maximum value difference between sample.
Since most of attributive character are intensive continuous in software defect data set, so selecting the Europe based on measurement several
Reed distance measures the dissimilarity between sample, to realize that software defect data set is empty from original feature space to dissimilarity
Between conversion.
(2) software defect is predicted.
When unknown sample arrives, calculates unknown sample first and prototype concentrates the dissimilarity between each sample, reflected
It is mapped to dissimilarity space；Then, using the software defect prediction model built to the unknown sample in dissimilarity space
This is predicted, you can learns whether unknown sample is defective.
The present invention is converted by prototype selection, dissimilarity and three links of classification form, Algorithms T-cbmplexity, that is, each
The summation of link time complexity.Give software defect imbalanced data sets D, sample size n, a number of attributes m, profit
With the present invention when carrying out failure prediction on D, the computational methods of each link time complexity are as described below：
1) time complexity of prototype selection：When choosing r prototype from the imbalanced data sets that sample size is n, no
The time complexity of same prototype selection method also differs.Random selection method sampling without replacement repeats r time complexity
For TRC=O (r).
2) time complexity of dissimilarity conversion：Dissimilarity conversion be intended to calculate original software defect data set with
Prototype concentrates the dissimilarity between sample, to convert it to dissimilarity space, the time complexity of dissimilarity conversion
For TDT=O (nr).
3) time complexity of failure prediction：It is n in sample size, number of attributes is r+1 based on the soft of dissimilarity
When being predicted on part defective data collection, time complexity depends on selected machine learning algorithm, TC=O (C (n, r+
1))。
By the above process it was determined that the time complexity T of the present inventionDSDPA=TRC+TDT+TC, i.e. TDSDPA=O (r)+O
(n·r)+O(C(n,r+1)).By the formula can be seen that using the program of the embodiment of the present invention on time complexity compared with
Low, realization is relatively easy to, and processing procedure is faster.
Comparative analysis of the embodiment of the present invention is using on the present invention and raw data set, arest neighbors (IB1), decision tree
(J48), 6 kinds of neural network (MLP), naive Bayesian (NB), random forest (RF) and support vector machines (SVM) machine learning sides
Estimated performance (AUC) of the method on 18 software defect data sets.Shown in table 1 specific as follows, compare for each algorithm performance.
Table 1
By table 1 as it can be seen that the present invention can to effectively improve conventional machines learning method pre- on software defect data set
Performance is surveyed, especially when carrying out software defect prediction using IB1 and support vector machines algorithm, IB1 algorithms are in software defect
Classification performance on data set averagely improves 11.11%；Classification performance of the SVM algorithm on software defect data set averagely carries
Rise 12%.Average classification performance of J48, MLP, NB algorithm on software defect data set is also improved, enhancing rate point
It Wei 3.12%, 2.5%, 2.6% and 2.7%.
In order to improve the discriminating power of attributive character in software defect data set, the embodiment of the present invention proposes a kind of by base
In the software defect prediction algorithm of dissimilarity, primitive attribute feature is substituted with dissimilarity between sample, not only remains data
Collect original statistical information, can also get in data set structural information, inherently improve to promote data set
The discriminating power of attributive character, and then improve learning performance of the conventional machines learning algorithm in software defect forecasting problem.
The embodiment of the present invention carries out prototype selection using random choice method, and Euclidean distance weighs the dissmilarity between sample
Property, 18 software defect data sets are transformed into dissimilarity space；Then, the embodiment of the present invention uses arest neighbors sorting algorithm
(1-NN, IB1), decision tree (DecisionTree, DT), neural network (Multi-layer Perceptron, MLP), simplicity
Bayes (Bayes, NB), random forest (Random Forest, RF) and support vector machines (Support Vector
Machine, SVM) 6 kinds of conventional machines learning algorithms, based on building prediction on the software defect data set in dissimilar space
Model；It was verified that the present invention can effectively improve the accuracy of software defect prediction.
The embodiment of the present invention is when assessing the performance of unbalanced learning method, using 10 × 10 folding cross validations, fully profit
While with data information, it is reduced as far as the accidental error of random sequence generation.
Although being example purpose, the preferred embodiment of the present invention is had been disclosed for, those skilled in the art will recognize
Various improvement, increase and substitution are also possible, and therefore, the scope of the present invention should be not limited to the above embodiments.
Claims (10)
1. a kind of Software Defects Predict Methods, which is characterized in that including：
The sample data for concentrating selection predetermined number in the first default initial data according to the first default selection rule, to obtain the
One prototype data collection；
It is calculated between the first default raw data set and the first prototype data collection not according to the first pre-determined distance algorithm
First data set of similitude；
Data in first data set are input to predetermined software bug prediction model, it is default original to obtain described first
The software defect prediction result of data set corresponding software, wherein shown predetermined software bug prediction model is according to default not phase
The model built like property.
2. Software Defects Predict Methods as described in claim 1, which is characterized in that according to the first default selection rule first
Default initial data concentrates the sample data of selection predetermined number, before obtaining the first prototype data collection, to further include：
The sample data for concentrating selection predetermined number in the second default initial data according to the second default selection rule, to obtain the
Diarch data set；
It is calculated between the second default raw data set and the second prototype data collection not according to the second pre-determined distance algorithm
Second data set of similitude；
According to the data and the default sorting algorithm structure predetermined software bug prediction model in second data set.
3. Software Defects Predict Methods as claimed in claim 1 or 2, which is characterized in that the first default selection rule or
The second default selection rule includes at least following one：Random selection method, the linear programming method based on cluster.
4. Software Defects Predict Methods as claimed in claim 1 or 2, which is characterized in that the first pre-determined distance algorithm or
The second pre-determined distance algorithm includes at least following one：Euclidean distance algorithm, manhatton distance algorithm, Min can husbands Si
Base distance algorithm.
5. a kind of software defect prediction meanss, which is characterized in that including：
First choice module, for concentrating selection predetermined number in the first default initial data according to the first default selection rule
Sample data, to obtain the first prototype data collection；
First computing module, for calculating the first default raw data set and described first according to the first pre-determined distance algorithm
First data set of dissimilarity between prototype data collection；
Prediction module, for the data in first data set to be input to predetermined software bug prediction model, to obtain
State the software defect prediction result of the first default raw data set corresponding software, wherein shown predetermined software bug prediction model
For according to the model for presetting dissimilarity structure.
6. software defect prediction meanss as claimed in claim 5, which is characterized in that further include：
Second selecting module, for concentrating selection predetermined number in the second default initial data according to the second default selection rule
Sample data, to obtain the second prototype data collection；
Second computing module, for calculating the second default raw data set and described second according to the second pre-determined distance algorithm
Second data set of dissimilarity between prototype data collection；
Module is built, for according to the data and the default sorting algorithm structure predetermined software defect in second data set
Prediction model.
7. such as software defect prediction meanss described in claim 5 or 6, which is characterized in that the first default selection rule or
The second default selection rule includes at least following one：Random selection method, the linear programming method based on cluster.
8. such as software defect prediction meanss described in claim 5 or 6, which is characterized in that the first pre-determined distance algorithm or
The second pre-determined distance algorithm includes at least following one：Euclidean distance algorithm, manhatton distance algorithm, Min can husbands Si
Base distance algorithm.
9. a kind of storage medium, is stored with computer program, which is characterized in that real when the computer program is executed by processor
Described in existing any one of claims 1 to 4 the step of Software Defects Predict Methods.
10. a kind of electronic equipment includes at least memory, processor, computer program, feature is stored on the memory
It is, the processor is realized soft described in any one of claims 1 to 4 in the computer program on executing the memory
The step of part failure prediction method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810162377.5A CN108647138A (en) | 2018-02-27 | 2018-02-27 | A kind of Software Defects Predict Methods, device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810162377.5A CN108647138A (en) | 2018-02-27 | 2018-02-27 | A kind of Software Defects Predict Methods, device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108647138A true CN108647138A (en) | 2018-10-12 |
Family
ID=63744085
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810162377.5A Pending CN108647138A (en) | 2018-02-27 | 2018-02-27 | A kind of Software Defects Predict Methods, device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108647138A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446090A (en) * | 2018-10-31 | 2019-03-08 | 南开大学 | Software defect prediction model based on deep neural network and Probabilistic Decision-making forest |
| CN111090579A (en) * | 2019-11-14 | 2020-05-01 | 北京航空航天大学 | Software defect prediction method based on Pearson correlation weighting association classification rule |
| CN113157564A (en) * | 2021-03-17 | 2021-07-23 | 江苏师范大学 | Cross-project defect prediction method based on feature distribution alignment and neighborhood instance selection |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105589806A (en) * | 2015-12-17 | 2016-05-18 | 北京航空航天大学 | SMOTE+Boosting algorithm based software defect tendency prediction method |
| CN105760889A (en) * | 2016-03-01 | 2016-07-13 | 中国科学技术大学 | Efficient imbalanced data set classification method |
| WO2017131263A1 (en) * | 2016-01-29 | 2017-08-03 | 한국과학기술원 | Hybrid instance selection method using nearest neighboring point for cross project defect prediction |
| CN107273387A (en) * | 2016-04-08 | 2017-10-20 | 上海市玻森数据科技有限公司 | Towards higher-dimension and unbalanced data classify it is integrated |
| CN107391369A (en) * | 2017-07-13 | 2017-11-24 | 武汉大学 | A kind of spanned item mesh failure prediction method based on data screening and data oversampling |
-
2018
- 2018-02-27 CN CN201810162377.5A patent/CN108647138A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105589806A (en) * | 2015-12-17 | 2016-05-18 | 北京航空航天大学 | SMOTE+Boosting algorithm based software defect tendency prediction method |
| WO2017131263A1 (en) * | 2016-01-29 | 2017-08-03 | 한국과학기술원 | Hybrid instance selection method using nearest neighboring point for cross project defect prediction |
| CN105760889A (en) * | 2016-03-01 | 2016-07-13 | 中国科学技术大学 | Efficient imbalanced data set classification method |
| CN107273387A (en) * | 2016-04-08 | 2017-10-20 | 上海市玻森数据科技有限公司 | Towards higher-dimension and unbalanced data classify it is integrated |
| CN107391369A (en) * | 2017-07-13 | 2017-11-24 | 武汉大学 | A kind of spanned item mesh failure prediction method based on data screening and data oversampling |
Non-Patent Citations (2)
| Title |
|---|
| 何中威 等: "《软件缺陷预测技术研究》", 《中国新通信》 * |
| 常瑞花 等: "《基于超欧氏距离近邻传播的软件缺陷预测方法》", 《计算机应用研究》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446090A (en) * | 2018-10-31 | 2019-03-08 | 南开大学 | Software defect prediction model based on deep neural network and Probabilistic Decision-making forest |
| CN111090579A (en) * | 2019-11-14 | 2020-05-01 | 北京航空航天大学 | Software defect prediction method based on Pearson correlation weighting association classification rule |
| CN113157564A (en) * | 2021-03-17 | 2021-07-23 | 江苏师范大学 | Cross-project defect prediction method based on feature distribution alignment and neighborhood instance selection |
| CN113157564B (en) * | 2021-03-17 | 2023-11-07 | 江苏师范大学 | Cross-project defect prediction method based on feature distribution alignment and neighborhood instance selection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113256066B (en) | PCA-XGboost-IRF-based job shop real-time scheduling method | |
| CN104798043B (en) | A kind of data processing method and computer system | |
| Sitompul et al. | Optimization model of K-Means clustering using artificial neural networks to handle class imbalance problem | |
| CN106537379A (en) | Fine-grained image similarity | |
| CN108647138A (en) | A kind of Software Defects Predict Methods, device, storage medium and electronic equipment | |
| CN111091177B (en) | Model compression method and device, electronic equipment and storage medium | |
| CN114841257B (en) | Small sample target detection method based on self-supervision comparison constraint | |
| CN111768285A (en) | Credit wind control model construction system and method, wind control system and storage medium | |
| JP2008527473A (en) | 3D model search method, search device, and search program | |
| CN110413775A (en) | A kind of data label classification method, device, terminal and storage medium | |
| WO2020012523A1 (en) | Information processing device, information processing method, and information processing program | |
| CN104392253A (en) | Interactive classification labeling method for sketch data set | |
| CN110210625A (en) | Modeling method, device, computer equipment and storage medium based on transfer learning | |
| CN109961093A (en) | A kind of image classification method based on many intelligence integrated studies | |
| CN108629345A (en) | Dimensional images feature matching method and device | |
| CN112632196A (en) | Data visualization method and device and storage medium | |
| Ghosh et al. | The class imbalance problem in deep learning | |
| CN109948825A (en) | Favorable Reservoir development area prediction technique based on improvement PSO in conjunction with Adaboost | |
| CN113762049B (en) | Content identification method, content identification device, storage medium and terminal equipment | |
| Shahbazi | Using decision tree classification algorithm to design and construct the credit rating model for banking customers | |
| CN114662989A (en) | Transient stability self-adaptive evaluation method and device for power system | |
| Banashree et al. | OCR for script identification of Hindi (Devnagari) numerals using feature sub selection by means of end-point with neuro-memetic model | |
| CN108073934A (en) | Nearly multiimage detection method and device | |
| CN110135769A (en) | Kinds of goods attribute fill method and device, storage medium and electric terminal | |
| CN114240069A (en) | Power grid online risk calculation method based on regression learning and feature mining |
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 |