CN111625441A - Unsupervised heterogeneous defect prediction method based on geodesic flow kernel - Google Patents
Unsupervised heterogeneous defect prediction method based on geodesic flow kernel Download PDFInfo
- Publication number
- CN111625441A CN111625441A CN201910144409.3A CN201910144409A CN111625441A CN 111625441 A CN111625441 A CN 111625441A CN 201910144409 A CN201910144409 A CN 201910144409A CN 111625441 A CN111625441 A CN 111625441A
- Authority
- CN
- China
- Prior art keywords
- geodesic
- domain
- source domain
- flow
- target
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000007547 defect Effects 0.000 title claims abstract description 30
- 238000007637 random forest analysis Methods 0.000 claims abstract description 11
- 238000006243 chemical reaction Methods 0.000 claims abstract description 7
- 230000002950 deficient Effects 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 19
- 238000000513 principal component analysis Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 238000005259 measurement Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims 1
- 238000012549 training Methods 0.000 abstract description 9
- 230000006978 adaptation Effects 0.000 abstract description 4
- 230000001131 transforming effect Effects 0.000 abstract 1
- 238000010276 construction Methods 0.000 description 15
- 238000013522 software testing Methods 0.000 description 5
- 239000000463 material Substances 0.000 description 2
- 244000298697 Actinidia deliciosa Species 0.000 description 1
- 235000009436 Actinidia deliciosa Nutrition 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an unsupervised heterogeneous defect prediction method based on geodesic flow-type kernels, which comprises the following steps: (1) transforming the optimal subspace dimension; (2) constructing a geodesic flow; (3) calculating a geodesic flow kernel; (4) building a balanced data set; (5) and (5) building a prediction model. The method introduces the Grassmann manifold into unsupervised field adaptation, considers the project data sets of a source domain and a target domain as two points in the Grassmann manifold, then constructs a geodesic flow between the two points, integrates all the points on the path along the flow to construct a geodesic path, and realizes the spatial conversion from the source domain and the target domain. And finally, training a random forest by using the converted source domain data to construct a defect prediction model to predict the defect tendency of the target project.
Description
Technical Field
The invention belongs to the technical field of software testing, in particular to a method for assisting software testers in improving software reliability, which is used for accurately judging defective software modules in software testing.
Background
With the continuous improvement of software scale, software projects are more and more complex, the problem of software quality becomes a focus of attention, and software companies need to invest in larger manpower and material resources for software testing. The software defect prediction technology can find and lock defective modules in advance by means of a machine learning method in the early stage of software development, so that resources can be reasonably distributed, and the quality of software is ensured.
Most of the defect prediction technologies mainly focus on the problem of defect prediction of the same project, namely, a prediction model is built based on a large amount of labeled data of the same project to predict an unmarked software model of the same project. However, in actual project development, especially newly developed software projects, there are not enough sets of tagged data, resulting in difficult model training. Fortunately, there are many project data sets currently available from open sources that can be used to build models for prediction, i.e., cross-project defect prediction techniques.
Most of cross-project defect prediction technologies require that a source project and a target project have the same measurement element, and there are many random factors in software development, such as differences in programming habits, proficiency, application fields, and the like of developers, which all cause differences in measurement indexes, measurement granularity, and the like of software projects, so that there is a problem of differential measurement indexes in software projects of different companies or data sets of different software projects of the same company. To solve the problem of metric element difference, heterogeneous defect prediction is required. Heterogeneous defect prediction employs tag data sets with different metric elements to find defective modules of a target item. The geodesic flow type kernel method is an unsupervised field adaptation method for transfer learning, and a source domain mapping and a target domain are mapped to a public space through a feature mapping, so that the distance between the source domain and the target domain is minimum.
In addition, the characteristics of the software data sets, such as the intrinsic measurement attributes of the software, are different, resulting in relatively large data variance and relatively small data variance. Moreover, for the software project, the twenty-eight law is met, i.e. a defective software module is much smaller than a non-defective software module (class imbalance problem). The class imbalance problem can bias the classifier towards non-defective modules, failing to resolve defective models, and degrading the performance of the classifier. The sampling method can adopt a random over-sampling method or a random under-sampling method to enable the data set to reach the balance.
Disclosure of Invention
In order to solve the problem of difference between measurement elements of a source project (source domain) and a measured project (target domain) in heterogeneous defect prediction and reduce the difference between the source domain and the target domain, the invention introduces Grassmann manifold and provides an unsupervised domain adaptation method of a geodesic flow kernel. And (3) regarding the project data sets of the source domain and the target domain as two points in the Grassmann flow, then constructing a geodesic flow between the two points, and integrating all the points on the path along the flow to construct a geodesic path so as to realize the spatial conversion from the source domain and the target domain. And finally, training a random forest by using the converted source domain data to construct a defect prediction model to predict the defect tendency of the target project. Specifically, the method comprises the following steps:
1) an optimal subspace dimension transform is selected. The data in the source domain S and the target domain T are first normalized so that all data are in the same range. Then, the preprocessed data are respectively subjected to feature transformation on the combination of a source domain, a target domain and a source domain and a target domain by using a principal component analysis method, and the converted source domain P is respectively calculatedSAnd a target domain PTTo the merged domain PS+TIs included angle of spaceAndminimization by greedy algorithmFinally, the optimal subspace dimension d of the principal component analysis is determined, so that the transformed feature set PSAnd PTIt is possible to express main information within the respective domains with the distance between the source domain and the target domain being minimized.
2) And constructing a geodesic line. Assuming the feature space P of the source domain and the target domain obtained in the step 1)SAnd PTMapping functions in the flow space by geodesic linesAfter mapping, at both poles 0 and 1, i.e.,. For a point t between two points from the source domain to the target domain in the manifold, the function value of the geodesic function can be expressed as:. In the formula, in the above-mentioned formula,is composed ofThe complement of (1), the dimension of which is D-D and satisfiesAnd = 0. To is pairPerforming singular value conversion to obtain、、Andnamely:,
3) and calculating a geodesic flow kernel. For smooth movement of the source domain to the target domain, the source domain is targetedAnd a target domainCalculate them by the idea of integration inUp-projection to an infinite-dimensional vector, the geodesic flow kernel is defined using the two infinite-dimensional projections, namely:
where G is a semi-positive definite matrix. The expression form of G is:
4) An equilibrium data set is established. And aiming at the G obtained in the step 3), the source domain and the target domain are transferred to the corresponding common space. And aiming at the condition that the number of defective examples in the source domain data is more than that of non-defective examples, sampling the defective examples by adopting a random oversampling method to obtain a data-balanced source domain data set.
5) And establishing a prediction model. And 4) establishing a classifier on the converted source domain data by using a random forest method according to the source domain data and the target domain data obtained in the step 4), predicting all samples of the target domain, and obtaining a result to calculate a corresponding performance evaluation index.
Further, the specific steps of the step 1) are as follows:
step 1) -1: a start state;
step 1) -2: respectively carrying out principal component analysis on the source domain data set S, the target domain data set T and the combined set S + T to obtain、And;
step 1) -3: computingAndincluded angle of space ofAndandincluded angle of space ofThereby obtaining;
Step 1) -4: repeating the steps 2 and 3 by adopting a greedy algorithm, thereby determining the optimal d;
step 1) -5: and finishing the dimension transformation of the optimal subspace.
Further, the specific steps of the step 2) are as follows:
step 2) -1: a start state;
step 2) -2: for the products obtained according to step 1)CalculatingIs not limited toSatisfy the following requirements=0;
Step 2) -5: and finishing the construction of the geodesic flow.
Further, the specific steps of the step 3) are as follows:
step 3) -1: a start state;
Step 3) -5: and finishing the geodesic flow core calculation.
Further, the specific steps of the step 4) are as follows:
step 4) -1: a start state;
Step 4) -3: after conversionRandom oversampling to obtain as many defective instances as non-defective ones
Further, the specific steps of the step 5) are as follows:
step 5) -1: a start state;
step 5) -2: computingObtained according to step 4)Training a random forest classifier to obtain a training model;
Step 5) -4: and finishing the establishment of the prediction model.
The invention discloses an unsupervised heterogeneous defect prediction method based on geodesic flow type kernel, which introduces Grassman manifold into unsupervised field adaptation, considers project data sets of a source domain and a target domain as two points in the Grassman manifold, then constructs a geodesic flow between the two points, integrates all the points on the path along the flow to construct a geodesic path, and realizes the spatial conversion from the source domain and the target domain. And finally, training a random forest by using the converted source domain data to construct a defect prediction model to predict the defect tendency of the target project. The method greatly improves the accuracy of prediction, enables software testers to reasonably distribute human and material resources, and improves the working efficiency of software testing, thereby effectively controlling the quality of software projects.
Drawings
Fig. 1 is a flowchart of an unsupervised heterogeneous defect prediction method based on a geodesic flow kernel in the implementation of the present invention.
Fig. 2 is a flow chart of the optimal subspace dimension transformation of fig. 1.
Fig. 3 is a flow chart of the geodesic construction of fig. 1.
Fig. 4 is a flow chart of the geodesic flow kernel calculation of fig. 1.
FIG. 5 is a flow chart of the balanced data set construction of FIG. 1.
FIG. 6 is a flow chart of the predictive model construction of FIG. 1.
Detailed Description
To further illustrate the technical content of the present invention, specific examples are described below with reference to the accompanying drawings.
Fig. 1 is a flowchart of an unsupervised heterogeneous defect prediction method based on a geodesic core according to an embodiment of the present invention.
An unsupervised heterogeneous defect prediction method based on geodesic flow kernel is characterized by comprising the following steps:
and S1, performing optimal subspace dimension transformation, and finding an optimal principal component analysis subspace d by using a greedy algorithm, so that the source domain and the target domain keep the maximum consistency under the d-dimensional subspace.
S2 geodesic flow construction, and according to the optimal subspace dimension d obtained by S1 construction, performing principal component analysis on the source domain and the target domain to obtainAndintroducing a Grassmann manifold,andrespectively mapping to two poles of 0 and 1, thereby obtaining the geodesic function of [0,1]A function value of a point t in between.
S3, calculating a geodesic flow kernel, obtaining the projection of the source domain vector and the target domain vector on the geodesic function by adopting the integral idea according to the geodesic function at the t obtained in S2, and defining the geodesic flow kernel according to the inner product of the projection.
S4 balance data set construction, converting the source domain and the target domain to a public space according to the G obtained in S3, copying a defective example of the converted source domain by adopting a random oversampling method, and balancing the defective example.
And S5 prediction model construction, namely obtaining balanced source domain label data according to S4, training a random forest classifier to obtain a prediction model with better performance, and predicting target domain data obtained in S4.
FIG. 2 is a flow chart of an optimal subspace dimension transformation. And finding the optimal principal component analysis subspace d by using a greedy algorithm. The method comprises the following specific steps.
Step 1: a start state; step 2: performing principal component analysis on the source domain, the target domain and the collection domain; and step 3: calculating included angles from the source domain and the target domain to the collection; and 4, step 4: obtaining an optimal subspace dimension d by adopting a greedy algorithm; and 5: and finishing the construction of the geodesic flow.
Fig. 3 is a flow chart of geodesic construction. And aiming at the obtained optimal subspace, introducing a Grassmann manifold to obtain a geodesic flow function. The method comprises the following specific steps.
Step 1: a start state; step 2: obtainedIs not limited to(ii) a And step 3: decomposition by SVGAnd(ii) a And 4, step 4: constructing a geodesic flow function; and 5: and finishing the construction of the geodesic flow.
FIG. 4 is a flow chart of computing a geodesic core. And calculating the geodesic current core by adopting an integral idea according to the geodesic function. The method comprises the following specific steps.
Step 1: a start state; step 2: calculating a space included angle; and step 3: calculating a matrix according to the included angle; and 4, step 4: calculating to obtain a geodesic current core; and 5: and finishing the geodesic current type nuclear calculation.
FIG. 5 is a flow chart of balanced data set construction. And sampling the source domain data by using a random oversampling method according to the geodesic flow type nuclear conversion source domain and the target domain. The method comprises the following specific steps.
Step 1: a start state; step 2: converting a source domain and a target domain; and step 3: randomly over-sampling source domain data; and 4, step 4: and finishing the construction of the balance data set.
FIG. 6 is a flow chart of predictive model construction. And training a random forest classifier according to the balanced source domain data set, and predicting a sample of the target domain by using the trained model. The method comprises the following specific steps.
Step 1: a start state; step 2: training a random forest classifier; and step 3: predicting a target domain instance; and 4, step 4: and providing the prediction result for a tester to perform test resource allocation.
In conclusion, the invention solves the problems of isomerism and class imbalance in defect prediction, not only improves the discovery of defect modules, but also improves the efficiency of software testing.
Claims (6)
1. An unsupervised heterogeneous defect prediction method based on geodesic flow type kernel is characterized in that a source domain project and a target domain project are regarded as two points in a Grassman flow, the geodesic flow is constructed, and the conversion from a source domain to a target domain space is realized; secondly, carrying out balance processing on the source domain project data by using a random over-adoption technology, and constructing a defect prediction model by combining a random forest method of machine learning to realize classification of the target project module; the method comprises the following steps:
1) selecting optimal subspace dimension transformation;
definition 1: the distance between the source domain and the target domain is expressed as the total measurement of the spatial included angle between the source domain and the merged domain, and the calculation formula is as follows:
merging the domains: a principal component analysis set of a merged set of source and target domains;
αd: the included angle between the d subspace and the merging domain after the source domain principal component analysis;
βd: the included angle between the d subspace and the merged domain space after the principal component analysis of the target domain;
finding an optimal principal component analysis subspace d by using a greedy algorithm, so that the source domain and the target domain keep the maximum consistency under the d-dimensional subspace;
2) constructing a geodesic line;
definition 1: the function value of the geodesic function at point t can be expressed as:
using the optimal subspace dimension d obtained in the step 1), carrying out principal component analysis on the source domain and the target domain to obtain PSAnd PTIntroduction of the Grassmann manifold, PSAnd PTRespectively mapping to two poles of 0 and 1, thereby obtaining the geodesic function of [0,1]A function value of a point t in between;
3) calculating a geodesic flow kernel;
definition 1: the geodesic flow kernel is defined using the two infinite-dimensional projections, namely:
wherein G is a semi-positive definite matrix, and the expression form of G is as follows:
Λ1,Λ2and Λ3Is a diagonal matrix, the elements of which are calculated as:
θiis PsAnd PTThe spatial included angle between the two parts;
using the geodesic function at the position t obtained in the step 2), adopting the integral idea to obtain the projection of the source domain vector and the target domain vector on the geodesic function, and defining a geodesic flow kernel according to the inner product of the projection;
4) establishing a balanced data set; using the G obtained in the step 3), transferring the source domain and the target domain to a corresponding common space, and sampling the defective examples by adopting a random oversampling method aiming at the condition that the defective examples in the source domain data are more than the non-defective examples to obtain a data-balanced source domain data set;
5) establishing a prediction model, using the source domain and target domain data obtained in the step 4), establishing a classifier on the converted source domain data by using a random forest method of machine learning, predicting all samples of the target domain, and obtaining a result to calculate a corresponding performance evaluation index.
2. The method for unsupervised heterogeneous defect prediction based on geodesic flow kernels according to claim 1, characterized in that in step 1) an optimal subspace dimension transformation is selected; and finding an optimal principal component analysis subspace d by using a greedy algorithm, so that the source domain and the target domain keep the maximum consistency under the d-dimensional subspace.
3. The unsupervised heterogeneous defect prediction method based on geodesic flow kernel of claim 1, characterized in that in step 2) geodesic lines are constructed; the source domain and target domain project data sets are considered as two points in a grassmann flow, and then a geodesic flow is constructed between the two points.
4. The unsupervised heterogeneous defect prediction method based on geodesic kernels according to claim 1, characterized in that in step 3) geodesic kernels are calculated; the method comprises the following specific steps: and (3) using the geodesic function at the position of the geodesic flow t, adopting the integral idea to obtain the projection of the vector of the source domain and the target domain on the geodesic function, and defining a geodesic flow kernel according to the inner product of the projection.
5. The method for unsupervised heterogeneous defect prediction based on geodesic flow kernels according to claim 1, characterized in that in step 4) a balanced dataset is established; the method comprises the following specific steps: and transferring the source domain and the target domain to a corresponding common space, and sampling the defective examples by adopting a random oversampling method aiming at the condition that the defective examples in the source domain data are more than the non-defective examples to obtain a data-balanced source domain data set.
6. The geodesic flow kernel based unsupervised heterogeneous defect prediction method of claim 1, characterized in that in step 5) a prediction model is established; and establishing a classifier on the converted source domain data by using a random forest method of machine learning, predicting all samples of the target domain, and calculating corresponding performance evaluation indexes according to results.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910144409.3A CN111625441A (en) | 2019-02-27 | 2019-02-27 | Unsupervised heterogeneous defect prediction method based on geodesic flow kernel |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910144409.3A CN111625441A (en) | 2019-02-27 | 2019-02-27 | Unsupervised heterogeneous defect prediction method based on geodesic flow kernel |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111625441A true CN111625441A (en) | 2020-09-04 |
Family
ID=72258752
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910144409.3A Pending CN111625441A (en) | 2019-02-27 | 2019-02-27 | Unsupervised heterogeneous defect prediction method based on geodesic flow kernel |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111625441A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112731285A (en) * | 2020-12-22 | 2021-04-30 | 成都中科微信息技术研究院有限公司 | Cross-time multi-source radio signal positioning method based on geodesic flow kernel transfer learning |
Citations (3)
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 |
CN105608004A (en) * | 2015-12-17 | 2016-05-25 | 云南大学 | CS-ANN-based software failure prediction method |
CN108563556A (en) * | 2018-01-10 | 2018-09-21 | 江苏工程职业技术学院 | Software defect prediction optimization method based on differential evolution algorithm |
-
2019
- 2019-02-27 CN CN201910144409.3A patent/CN111625441A/en active Pending
Patent Citations (3)
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 |
CN105608004A (en) * | 2015-12-17 | 2016-05-25 | 云南大学 | CS-ANN-based software failure prediction method |
CN108563556A (en) * | 2018-01-10 | 2018-09-21 | 江苏工程职业技术学院 | Software defect prediction optimization method based on differential evolution algorithm |
Non-Patent Citations (1)
Title |
---|
BOQING GONG: "Geodesic Flow Kernel for Unsupervised Domain Adaptation" * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112731285A (en) * | 2020-12-22 | 2021-04-30 | 成都中科微信息技术研究院有限公司 | Cross-time multi-source radio signal positioning method based on geodesic flow kernel transfer learning |
CN112731285B (en) * | 2020-12-22 | 2023-12-08 | 成都中科微信息技术研究院有限公司 | Cross-time multi-source radio signal positioning method based on geodesic flow kernel migration learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | TLEL: A two-layer ensemble learning approach for just-in-time defect prediction | |
Jing et al. | An improved SDA based defect prediction framework for both within-project and cross-project class-imbalance problems | |
Kluza et al. | Innovation and environmental, social, and governance factors influencing sustainable business models-Meta-analysis | |
Berger et al. | A volumetric deep convolutional neural network for simulation of mock dark matter halo catalogues | |
Kasiviswanathan et al. | Methods used for quantifying the prediction uncertainty of artificial neural network based hydrologic models | |
Hill et al. | Model-independent method for measuring the angular coefficients of B0→ D∗− τ+ ντ decays | |
Lakshminarayana et al. | Automatic generation and optimization of test case using hybrid cuckoo search and bee colony algorithm | |
Yuan et al. | ALTRA: Cross-project software defect prediction via active learning and tradaboost | |
Louzada et al. | Poly-bagging predictors for classification modelling for credit scoring | |
CN111161249B (en) | Unsupervised medical image segmentation method based on domain adaptation | |
Yu et al. | Robust calibration with multi-domain temperature scaling | |
Hasnain et al. | Benchmark dataset selection of Web services technologies: A factor analysis | |
Gouveia et al. | A full ARMA model for counts with bounded support and its application to rainy-days time series | |
Castaño et al. | Exploring the Carbon Footprint of Hugging Face's ML Models: A Repository Mining Study | |
Chen et al. | Transportation infrastructure and economic growth in China: A meta-analysis | |
Zhang et al. | Feature relevance term variation for multi-label feature selection | |
CN115423594A (en) | Enterprise financial risk assessment method, device, equipment and storage medium | |
CN111625441A (en) | Unsupervised heterogeneous defect prediction method based on geodesic flow kernel | |
Angeloudi et al. | ERGO-ML: towards a robust machine learning model for inferring the fraction of accreted stars in galaxies from integral-field spectroscopic maps | |
CN112860531B (en) | Block chain wide consensus performance evaluation method based on deep heterogeneous graph neural network | |
Kipkogei et al. | Business success prediction in Rwanda: a comparison of tree-based models and logistic regression classifiers | |
CN116956105A (en) | Classification model training method, defect identification method, device and electronic equipment | |
Lakra et al. | Improving software maintainability prediction using hyperparameter tuning of baseline machine learning algorithms | |
Castaño Fernández | A greenability evaluation sheet for ai-based systems | |
Li et al. | Sentence dependent-aware network for aspect-category sentiment analysis |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200904 |
|
RJ01 | Rejection of invention patent application after publication |