CN108614778B - Android App program evolution change prediction method based on Gaussian process regression - Google Patents
Android App program evolution change prediction method based on Gaussian process regression Download PDFInfo
- Publication number
- CN108614778B CN108614778B CN201810443100.XA CN201810443100A CN108614778B CN 108614778 B CN108614778 B CN 108614778B CN 201810443100 A CN201810443100 A CN 201810443100A CN 108614778 B CN108614778 B CN 108614778B
- Authority
- CN
- China
- Prior art keywords
- change
- class
- version
- matrix
- index
- 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.)
- Active
Links
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
- G06F11/3668—Software testing
- G06F11/3672—Test management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Stored Programmes (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the field of software engineering technology and application, and provides a more detailed version evolution result for a project manager and a developer in order to finally output a variable category sorting list about the category change size, thereby saving the test time and the test cost. The invention discloses an android App program evolution change prediction method based on Gaussian process regression, which comprises the following steps: crawling data in a first stage; preprocessing data in a second stage; stage three, obtaining the size of an optimal prediction window; and step four, constructing a Gaussian process regression prediction model, and finally outputting a next version class change degree ranking list through the prediction model. The invention is mainly applied to the technical field of software engineering.
Description
Technical Field
The invention relates to the field of software engineering technology and application, in particular to a Gaussian process regression-based Android App (Android application) program evolution change prediction method.
Background
The real world is constantly changing and developing, and when the system environment changes, new demands will appear. To accommodate the changing environmental needs, software systems need to be continuously maintained and evolved. Software evolution is therefore a process by which programs are continually tuned to meet new software requirements. However, on the one hand: software maintenance costs typically exceed 50% of the total software lifecycle cost, and volatile classes in open-source products take much time and effort on the part of developers, which also increases project costs. On the other hand: the Pareto's Law is verified to be true in software engineering, and most (80%) software changes are derived from a small (about 20%) class, which is important and is the source of defects and changes.
We need to identify volatile classes to reduce software maintenance costs, and currently, much research is devoted to help developers identify key and low quality parts in software systems, namely: a class that tends to be changed is identified. Determining the changeable classes of the software by developing a prediction model is an important research field, and 1) taking the classes to be changed in the next version as the starting points of software testing, preferentially and deeply testing, effectively reconstructing and designing the changeable classes, and reducing software faults. 2) The method helps developers and project managers to effectively distribute limited resources, thereby reducing development time, energy and cost.
In the process of implementing the invention, the inventor finds that the prior art has at least the following disadvantages:
(1) statistics between head-to-tail and mid-segment versions are established: most of them are built between the initial version and the final version of evolution, count metrics (index group for measuring code characteristics) of the initial version, and count which classes (classes in java program) are changed; or, only the first and the last two versions are counted by using several evolution processes in the whole evolution process.
(2) Regarding the selected Metrics: they do not describe the entire evolution process with metrics, the selected metric being a measure of class in the first version. One author proposes an index based on evolution, and although all evolution processes are counted, only the change times, the first change version number and the last change version number are calculated, and no more detailed evolution process is calculated, so that the counting is too coarse.
(3) And (3) a classification algorithm: the authors use the classification concept, metric as the independent variable and binary change/invariance as the dependent variable, to construct a classification model of change tendency classes. Only can identify whether the class is variable in the next version, but can not give the change degree, so as to better guide a developer to predict the version evolution result.
Aiming at the problems, the invention provides a method for predicting the class with the change tendency of the newly released version of the Android App. And complete increment information in the evolution process is collected, and the change degree of the class in the next version is predicted based on a Gaussian process regression model and the size of a prediction window, so that the time, the energy and the cost of software maintenance are reduced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a prediction method for the variable change degree of the Android App. Under the framework, a user provides a mobile phone application program App to be predicted, the system firstly crawls App evolution historical data, and complete metadata and source code information between every two continuous versions are converted into an incremental matrix model based on time slices. Along with the regression prediction model of the Gaussian process and the size of the optimal prediction window, a classification-variable ordered list related to the size of class change is finally output, a more detailed version evolution result can be obtained for project managers and developers, and the test time and the test cost are saved. Therefore, the technical scheme adopted by the invention is that the android App program evolution change prediction method based on Gaussian process regression comprises the following steps:
stage one data crawling: crawling evolution history data through an open source software version control system application programming interface GithubApi, mining complete evolution history information, and analyzing the complete evolution history to obtain an evolution rule;
and (3) preprocessing data in a second stage: two new types of indexes are adopted, class change indexes are used for researching the problem of version change prediction by class change metrics and change degree index change degree metrics, which classes are to be changed in newly issued versions, and in order to obtain the result, complete metadata and source code information between two continuous versions are converted into an increment matrix model based on time slices;
stage three, obtaining the optimal prediction window size: window size: the number of versions to predict that can result in the best prediction result; determining the weight occupied by each evolution matrix according to the influence of the evolution matrix on the subsequent version, and obtaining fixed weights corresponding to different matrixes in the window when the predicted window size is different by using an entropy weight method and the information entropy provided by the evolution matrices;
and step four, constructing a Gaussian process regression prediction model: according to the selected optimal window size m, selecting m evolution increment matrixes as independent variables, predicting the change degree of a certain class as dependent variables, expanding the matrixes according to metrics columns, constructing a prediction model by adopting a Gaussian process regression algorithm, processing data of high-dimensional small samples by adopting a Gaussian process, and finally outputting a next version class change degree ordered list through the prediction model.
The method also comprises the following steps:
a scanning module: extracting all class from the first version to the last released version of the App project;
extracting each class index metadata: for two continuous versions, extracting two types of metadata information of each class and calculating the code similarity of the common class, and marking each class by adopting an inter-version variation measurement element, wherein three types of indexes are described as follows:
class i change metric index: the change information of class at the file level between two continuous versions is described, namely whether the file is newly added, deleted or modified in the next version;
class ii change degree measurement index: describing the inter-version distance of the class between two consecutive versions using a value between 0 and 1;
classical object-oriented index iii: and describing change information between versions by using the object-oriented index difference value. Compared with the indexes, the method has the advantages that the change of each class is described in a finer granularity, the class is marked by the classical object-oriented indexes, and the indexes comprise an index group CK, an index group QMOOD and an index group Ca Ce indexes which are jointly proposed by Chidammer S and KemerC and are used for describing object-oriented features of the class, so that the difference value of the indexes of the common class between two versions between the versions is obtained;
(3) generating and storing: an evolving incremental history matrix is integrated every two adjacent versions, and the series of matrices are stored in a database.
Step two training and prediction
Based on the incremental matrix model obtained in the first stage, obtaining matrix weight through an entropy weight method, training a matrix through a machine learning algorithm to obtain the optimal prediction window size, finally, predicting the class change condition of a newly released version based on a Gaussian process regression model according to the optimal prediction window size of a corresponding App, and outputting a classification-variable change ranking list related to the change size; calculating the matrix information entropy in each group of matrixes, wherein the number of the matrixes in each group is m, and further determining the weight of each matrix in the window m; the software evolves a total of n versions, so that n-1 matrix incremental matrix models can be obtained 2 ～matrix n Will matrix 2 ～matrix n The total number of the effective change information is taken as the statistics of the matrix characteristics, i.e. addend, delete and modifyThe indexes respectively describe whether a certain class is a new class in a new version, a deleted class in the new version, a modified class in the new version and not the sum of the information number of 0, so that n-m index groups exist.
Determining the optimal prediction window size:
data processing: and (4) standardizing the index data.
Solving the information entropy of each index group: calculating the information entropy of each matrix according to the definition of the information entropy in the information theory;
determining the weight of each index: calculating the weight of each matrix through the information entropy;
obtaining the optimal prediction window size: and constructing multiple logistic regression models according to different prediction window sizes, wherein the fitting effect of the models determines the window size to predict the next version of the App, and the evolution change data of the previous version is used for predicting the change of a class of the next version.
The invention has the characteristics and beneficial effects that:
the invention provides a method for predicting the variable change degree based on the Gaussian process by mining complete version evolution history change information. Compared with the prior art, the method has the advantages that a more optimized prediction effect and a more detailed prediction result are obtained, the change tendency is predicted, the change degree of the variable class is further predicted, the ordered list of the change degree of the variable class is output, and the App is conveniently maintained and managed by a project manager and a developer. The expected benefits include:
1) correct classification ratio of volatile classes for newly released versions: compared with other prediction methods, the model is constructed based on the App complete increment information as an index, and more accurate and correct classification rate can be obtained. And (4) classification: the class is "with a tendency to change" and the class is "without a tendency to change".
2) Predicting the stability of the changeful method of the new release version: and for any App, stable prediction accuracy can be obtained, and universality is strong.
3) The invention provides an accuracy of a sorted list of the degree of variable changes: the category with large variation indicates that the prediction accuracy of the sorted list is better at the top of the list, and the accuracy of the ranked list is evaluated through a mean evaluation index MAP (i.e., average accuracy, a sort evaluation index) and an NDGC (normalized discrete cumulative gain, a sort evaluation index). Three bivariate correlation analysis methods, namely Pearson correlation analysis, Spirann correlation analysis, Kandler grade correlation analysis, comparison of predicted class ranking and real class ranking, and comparison of the correlation of predicted class alteration degree and real class alteration degree are used.
Description of the drawings:
FIG. 1 is an overall framework diagram of the new version changeability prediction of the present invention.
Fig. 2 is a list of apps used in the test according to the present invention.
FIG. 3 is a framework diagram of the data crawling phase of the present invention.
FIG. 4 is a list of classical object-oriented metrics according to the present invention.
FIG. 5 is a schematic diagram of the incremental matrix model of the present invention.
FIG. 6 is a block diagram of a data preprocessing framework according to the present invention.
FIG. 7 is a diagram illustrating an index set obtained by the entropy weight method according to the present invention.
FIG. 8 is a block diagram of the acquisition prediction window size according to the present invention.
FIG. 9 is a schematic diagram of a regression model of a Gaussian process constructed according to the present invention.
Detailed Description
The invention provides a method for crawling and preprocessing historical information of software evolution change, which is based on a Gaussian process regression model and the size of an optimal prediction window, excavates potential rules in an evolution process, predicts which classes in an Android App will be changed in a newly-released version, and finally obtains a change ranking list according to the change size of the classes. Therefore, the class at the front end of the list can be subjected to priority test and depth test, and the test time and cost are saved.
The invention provides a method for predicting the changeability degree of Android App, which predicts the changeability degree by evolution historical information, wherein a prediction frame is divided into two main parts, namely a data collection stage and a training and prediction stage. The data collection stage comprises two sub-stages, namely data crawling and data preprocessing; the training and predicting part comprises two sub-stages, and the optimal prediction window size and regression model construction are found. And finally outputting a sorted list of the variable change degrees of the next version for the user.
The overall execution flow of the invention method is as follows:
stage one data crawling: the evolution history data is crawled through a Github Api (Github is an open source software version control system, and the Github Api is an application programming interface). And (4) excavating complete evolution history information, and obtaining an evolution rule by analyzing the complete evolution history.
And (3) preprocessing data in a second stage: and integrating evolution historical data to obtain an incremental matrix model between two continuous versions. The next version of the software is generated based on the previous version. For a class in the object-oriented system, the time slice-based incremental information, that is, the change of the class and the addition, deletion and modification of the related class, may become the reasons for the defect and the introduction of the change in the new version, and is a good index in the potential object-oriented system. Therefore, in this paper, we propose two new classes of metrics, class change metrics & change degree metrics, to study the version change prediction problem, which classes will change in the newly published version. To achieve this result, the complete metadata and source code information between two successive versions is translated into a time slice based delta matrix model.
Stage three, obtaining the optimal prediction window size: the evolution result of the next version may be related to the previous evolution versions. Window size: the number of versions to predict that can result in the best prediction result. The strength of the correlation with the prediction result is judged by selecting the prediction result when the window sizes are different, and the window size m corresponding to the model with the best fitting degree is the optimal prediction window size. However, the influence of the previous versions on the next version may be different, and the version closest to the current version may have the greatest influence on the prediction result, but these conditions may not have a great difference in the degree of influence on the prediction result, so that experimental judgment is also needed. Therefore, before prediction, in order to obtain a more accurate prediction result, the weight occupied by each evolution matrix should be determined according to the influence of the evolution matrix on subsequent versions. The fixed weights corresponding to different matrixes in the window when the size of the prediction window is different are obtained through the information entropy provided by the evolution matrixes by adopting an entropy weight method.
And step four, constructing a Gaussian process regression prediction model: and according to the selected optimal window size m, selecting m evolution increment matrixes as independent variables, and taking the predicted change degree of a certain class as a dependent variable. The number of versions of the evolution data is probably from 6 to 62, but the number of the prediction indexes is large, the number of classes is large, the matrix is expanded according to metrics columns, so that the index dimension is large, a Gaussian process regression algorithm is selected to construct a prediction model, and the Gaussian process can process the data of high-dimensional small samples. And finally, outputting a next version class change degree ordered list through a prediction model.
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The invention provides a method for predicting the variable class change degree based on a Gaussian process. The general flow chart is shown in FIG. 1 and described in detail in the following
Step one data collection
The data collection phase is the first main phase of the variable prediction, and historical version information and historical modification information are crawled from a popular open source software development and hosting platform such as GitHub, so that quantitative analysis and maintenance management can be performed on some android applications. And storing the information into a local MySql database, and then performing data preprocessing on the obtained evolution historical data to obtain an increment matrix model based on time slices between every two continuous versions. Experiments were performed on 17 Android apps in three categories, and the selected apps and their descriptions are shown in fig. 2.
Data crawling:
(1) and acquiring description information of the App project, Apk (android installation package), history submission information and downloading source code information of all versions to the local through the Github Api.
(2) And (4) storing the original data into a database by using the class name and the class source code extracted from the source code.
(3) And obtaining complete evolution data information according to the version: function description information, version number, version name, historical submission information (which files are modified, the number of code lines modified, change description, etc.), class name, and source code information corresponding to each class can be queried from the database. The specific flow of data crawling is shown in fig. 3.
Data preprocessing:
a scanning module: and extracting all classes appearing from the first version to the last released version of the App item, and finally obtaining a union of the classes contained in all the versions.
Extracting each class index metadata: for two consecutive versions, two types of metadata information for each class are extracted and the code similarity for common classes is calculated. Each class is marked with 22 inter-version change metrics. Three types of indicators are described as follows:
class i change metric index: class change information at the file level between two successive versions is described, i.e., whether the file is newly added, deleted, modified in the next version.
Class ii change degree measurement index: the inter-version distance (degree of change) of the class between two consecutive versions is described using a value between 0 and 1. After blanks and comments are removed from the source code, the code similarity is calculated in a mode of the longest public subsequence, and 1 minus the code similarity is used, so that the change degree of a certain class between two versions is obtained. Therefore, the objective is to directly give a quantitative measure of the class change result.
III classical object-oriented index: and describing change information between versions by using the object-oriented index difference value. Each class change is described at a finer granularity than the previous indexes, and is labeled with classical object-oriented indexes, including CK (index set describing class object-oriented features commonly proposed by chidammber S and kemer C), QMOOD (index set proposed by j.bansiya and c.g.davis), and Ca Ce index (index set proposed by r.c.matrix), resulting in a difference in indexes between the two versions that share class. Thus, the object-oriented feature of class change is described more finely. The index used is described in figure 4.
(3) Generating and storing: thus, an evolving incremental history matrix can be integrated every two adjacent versions, and the series of matrices are stored in the database. The rows of the incremental matrix represent the classes contained by the system and the columns of the matrix represent the values of the contained indices. Wherein the class change index is 0 or 1, the class change degree index is a number between 0 and 1, and the object-oriented index difference is determined after calculation according to different object-oriented indexes. An example of the incremental matrix model is shown in FIG. 5. The complete data pre-processing flow is shown in fig. 6.
Step two training and prediction
The training and predicting stage is the second main stage of variable prediction, based on the incremental matrix model obtained in the first stage, the matrix weight is obtained through an entropy weight method, and the optimal prediction window size is obtained through a machine learning algorithm training matrix. And finally, predicting the class change condition of the newly released version based on a Gaussian process regression model according to the optimal prediction window size of the corresponding App, and outputting a variable class change ordered list related to the change size. And calculating the matrix information entropy in each group of matrixes (the number of the matrixes in each group is m), and further determining the weight of each matrix in the window m. The matrix with larger weight calculated from the m matrixes provides more information, and has large influence on the prediction matrix. Assuming that the software evolves for a total of n versions, n-1 matrices are obtained 2 ～matrix n In the experiment, matrix is used 2 The total number of effective change information contained in matrixn is taken as the statistics of matrix characteristics, namely the total number of information with addition, deletion and modification not being 0, which reflects the information amount provided by the matrix for the next version prediction to a certain extent, x 1 ,x 2 ,x 3 …x n . Assuming a window size of m, there are n-m index sets (e.g., m is 3, version 2, version 3, version 4 predicts version 5, version 3, version 4, version 5 predicts version 6 …, and so on, and version 2, 3, 4, version 3, 4, 5, version 4, 5, 6 … is one index set), and for a clearer description, each index set extracted by n-1 matrices is shown in fig. 7.
Determining the optimal prediction window size:
the flow chart for determining the optimal prediction window size is shown in fig. 8.
Data processing: and (4) standardizing the index data.
Solving the information entropy of each index group: and calculating the information entropy of each matrix according to the definition of the information entropy in the information theory.
Determining the weight of each index: and calculating the weight of each matrix through the information entropy.
Obtaining the optimal prediction window size: and constructing multiple logistic regression models according to different prediction window sizes, wherein the fitting effect of the models determines the window size to predict the next version of the App, and the evolution change data of the previous version is used for predicting the change of a class of the next version. And selecting the size of the prediction window corresponding to the model with the highest fitting degree as the size of the optimal prediction window, and constructing a regression model.
Constructing a regression model:
the regression model was constructed as shown in FIG. 9.
And selecting a Gaussian process regression algorithm to construct a prediction model, wherein the Gaussian process regression is suitable for the data of the high-dimensional small samples. The number of historical versions of the Android App may be tens to hundreds, but the number of classes contained in the source code is huge. Thousands of classes, indicating large data dimensions; we predict the following version change case with the previous version information, which belongs to the case of small samples because the number of versions is small. Then, according to the selected optimal window size m, selecting m evolution increment matrixes as independent variables, taking the predicted change degree of a certain class as a dependent variable, and constructing a Gaussian process regression model to obtain the change degree of the class.
Claims (3)
1. A prediction method for android App program evolution change based on Gaussian process regression is characterized by comprising the following steps:
stage one data crawling: crawling evolution history data through an open source software version control system application programming interface GithubApi, mining complete evolution history information, and analyzing the complete evolution history to obtain an evolution rule;
and (3) preprocessing data in a second stage: two new types of indexes, namely a class change index class change metrics and a change degree index change degree metrics are adopted to research the version change prediction problem, wherein the change degree index is a change degree metric index of a II type: describing the inter-version distance of the class between two consecutive versions using a value between 0 and 1;
which classes are to be changed in the newly published version, and in order to get this result, the complete metadata and source code information between two successive versions is converted into a slice-based incremental matrix model;
stage three, obtaining the optimal prediction window size: window size: the number of versions to predict that can result in the best prediction result; determining the weight occupied by each evolution increment matrix according to the influence of the evolution increment matrix on the subsequent version, and obtaining fixed weights corresponding to different matrixes in the window when the predicted window size is different by using an entropy weight method and the information entropy provided by the evolution increment matrices;
and step four, constructing a Gaussian process regression prediction model: according to the selected optimal window size m, selecting m evolution increment matrixes as independent variables, predicting the change degree of a certain class as dependent variables, expanding the matrixes according to metrics columns, constructing a prediction model by adopting a Gaussian process regression algorithm, processing data of high-dimensional small samples by adopting a Gaussian process, and finally outputting a next version class change degree ordered list through the prediction model.
2. The android App program evolution change prediction method based on gaussian process regression as claimed in claim 1, further comprising data preprocessing:
a scanning module: extracting all class from the first version to the last released version of the App project;
extracting each class index metadata: for two continuous versions, extracting two types of metadata information of each class and calculating the code similarity of the common classes, and marking a change measurement index between the versions of each class, wherein the three types of indexes are described as follows:
class i change metric index: the change information of class at the file level between two continuous versions is described, namely whether the file of the previous version is newly added, deleted or modified in the next version;
class ii change degree measurement index: describing the inter-version distance of the class between two consecutive versions using a value between 0 and 1;
III classical object-oriented index: describing change information between versions by using the object-oriented index difference value, describing the change of each class in a finer granularity compared with the previous indexes, and marking the class by using the classical object-oriented indexes, wherein the class comprises indexes of a description class object-oriented characteristic-oriented index group CK, an index group QMOOD and an index group Ca Ce which are jointly proposed by Chidamer S and Kemer C, and the difference value of the indexes of the common class between the two versions between the versions is obtained;
generating and storing: each two adjacent versions are integrated into an evolving delta matrix and the series of matrices are stored in a database.
3. The gaussian process regression-based android App program evolution change prediction method of claim 1, wherein training and predicting:
based on the incremental matrix model obtained in the second stage, obtaining matrix weight through an entropy weight method, training a matrix through a machine learning algorithm to obtain the optimal prediction window size, finally, predicting the class change condition of the newly released version based on a Gaussian process regression model according to the optimal prediction window size of the corresponding App, and outputting a classification change ranking list which is easy to change and is related to the change size; computing moments in each set of matricesThe entropy of the array information is m, and the weight of each matrix in the window m is further determined; software evolves a total of n versions, so that n-1 matrixes matrix can be obtained 2 ～matrix n Will matrix 2 ～matrix n The total number of the contained effective change information is taken as the statistics of the matrix characteristics, so that n-m index groups exist;
determining the optimal prediction window size:
data processing: standardizing each index data;
solving the information entropy of each index group: calculating the information entropy of each matrix according to the definition of the information entropy in the information theory;
determining the weight of each index: calculating the weight of each matrix through the information entropy;
obtaining the optimal prediction window size: and constructing multiple logistic regression models according to different prediction window sizes, wherein the fitting effect of the models determines the window size to predict the next version of the App, and the evolution change data of the previous version is used for predicting the change of a class of the next version.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810443100.XA CN108614778B (en) | 2018-05-10 | 2018-05-10 | Android App program evolution change prediction method based on Gaussian process regression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810443100.XA CN108614778B (en) | 2018-05-10 | 2018-05-10 | Android App program evolution change prediction method based on Gaussian process regression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108614778A CN108614778A (en) | 2018-10-02 |
| CN108614778B true CN108614778B (en) | 2022-08-16 |
Family
ID=63662737
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810443100.XA Active CN108614778B (en) | 2018-05-10 | 2018-05-10 | Android App program evolution change prediction method based on Gaussian process regression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108614778B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109714324B (en) * | 2018-12-18 | 2021-06-22 | 中电福富信息科技有限公司 | User network abnormal behavior discovery method and system based on machine learning algorithm |
| CN111290777B (en) * | 2020-01-23 | 2021-09-17 | 复旦大学 | Code change sequence method oriented to software code unit and code measurement |
| CN113762589A (en) * | 2021-07-16 | 2021-12-07 | 国家电网有限公司 | Power transmission and transformation project change prediction system and method |
| CN114037284B (en) * | 2021-11-10 | 2024-05-28 | 天津大学 | Method for predicting App popularity evolution result based on multi-layer attribute network |
| CN116303026B (en) * | 2023-03-20 | 2024-01-23 | 安徽工业大学 | Evolution history-based software architecture degradation prediction method, device and equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101908020A (en) * | 2010-08-27 | 2010-12-08 | 南京大学 | Method for prioritizing test cases based on classified excavation and version change |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| LT3197164T (en) * | 2011-10-17 | 2021-02-10 | Kabushiki Kaisha Toshiba | Decoding apparatus and decoding method |
| CN102694800A (en) * | 2012-05-18 | 2012-09-26 | 华北电力大学 | Gaussian process regression method for predicting network security situation |
| CN105302724A (en) * | 2015-11-05 | 2016-02-03 | 南京大学 | Instant defect predicting method based on mixed effect removing |
| US10084645B2 (en) * | 2015-11-30 | 2018-09-25 | International Business Machines Corporation | Estimating server-change risk by corroborating historic failure rates, predictive analytics, and user projections |
| CN106201871B (en) * | 2016-06-30 | 2018-10-02 | 重庆大学 | Based on the Software Defects Predict Methods that cost-sensitive is semi-supervised |
| CN106528428B (en) * | 2016-11-24 | 2019-06-25 | 中山大学 | A kind of construction method of software mutability prediction model |
-
2018
- 2018-05-10 CN CN201810443100.XA patent/CN108614778B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101908020A (en) * | 2010-08-27 | 2010-12-08 | 南京大学 | Method for prioritizing test cases based on classified excavation and version change |
Non-Patent Citations (1)
| Title |
|---|
| 基于演化数据的软件缺陷预测性能改进;王丹丹等;《软件学报》;20161215(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108614778A (en) | 2018-10-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108614778B (en) | Android App program evolution change prediction method based on Gaussian process regression | |
| US10360517B2 (en) | Distributed hyperparameter tuning system for machine learning | |
| CN113779272A (en) | Data processing method, device and equipment based on knowledge graph and storage medium | |
| CN112150237B (en) | Multi-model fused order overdue early warning method, device, equipment and storage medium | |
| CN112910690A (en) | Network traffic prediction method, device and equipment based on neural network model | |
| CN113221960B (en) | Construction method and collection method of high-quality vulnerability data collection model | |
| CN115641162A (en) | Prediction data analysis system and method based on construction project cost | |
| CN110706015A (en) | Advertisement click rate prediction oriented feature selection method | |
| CN110310012B (en) | Data analysis method, device, equipment and computer readable storage medium | |
| CN110689368A (en) | Method for designing advertisement click rate prediction system in mobile application | |
| CN113537807A (en) | Enterprise intelligent wind control method and device | |
| CN111582488A (en) | Event deduction method and device | |
| CN113672506B (en) | Dynamic proportion test case sorting and selecting method and system based on machine learning | |
| CN113177644A (en) | Automatic modeling system based on word embedding and depth time sequence model | |
| CN117172549A (en) | Construction budget generation method, construction budget generation system, terminal equipment and storage medium | |
| CN115238583B (en) | Business process remaining time prediction method and system supporting incremental log | |
| CN111324594A (en) | Data fusion method, device, equipment and storage medium for grain processing industry | |
| CN116341059A (en) | Tunnel intelligent design method based on similarity | |
| CN115630732A (en) | City operation-oriented enterprise migration big data monitoring and early warning method and device | |
| CN115616408A (en) | Battery thermal management data processing method and system | |
| CN115238965A (en) | Technical trend prediction method, device, equipment and medium based on patent information | |
| CN111008038B (en) | Pull request merging probability calculation method based on logistic regression model | |
| CN115617790A (en) | Data warehouse creation method, electronic device and storage medium | |
| CN112907055A (en) | Data processing aging test method and device | |
| CN117076454B (en) | Engineering quality acceptance form data structured storage method and system |
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 |