CN111858328A - Software defect module severity prediction method based on ordered neural network - Google Patents

Software defect module severity prediction method based on ordered neural network Download PDF

Info

Publication number
CN111858328A
CN111858328A CN202010679603.4A CN202010679603A CN111858328A CN 111858328 A CN111858328 A CN 111858328A CN 202010679603 A CN202010679603 A CN 202010679603A CN 111858328 A CN111858328 A CN 111858328A
Authority
CN
China
Prior art keywords
neural network
defect
network model
severity
software
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010679603.4A
Other languages
Chinese (zh)
Other versions
CN111858328B (en
Inventor
陈翔
贾焱鑫
李春明
葛骅
杨光
林浩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nantong University
Original Assignee
Nantong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nantong University filed Critical Nantong University
Priority to CN202010679603.4A priority Critical patent/CN111858328B/en
Publication of CN111858328A publication Critical patent/CN111858328A/en
Application granted granted Critical
Publication of CN111858328B publication Critical patent/CN111858328B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06F11/3604—Software analysis for verifying properties of programs
    • 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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • Quality & Reliability (AREA)
  • Stored Programmes (AREA)

Abstract

The invention belongs to the field of software quality assurance, and discloses a software defect module severity degree prediction method based on an ordered neural network. The method provided by the invention comprises the following steps: mining a version control system and a defect tracking system where a sample software project is located, marking a program module of the sample software project with measurement and defect severity, and constructing a sample data set; based on the sample data set, adopting an ordered neural network model and a Bayesian hyperparameter optimization method to obtain a software defect prediction model, namely the ordered neural network model with the optimal hyperparameters; and predicting the severity of the defects of the program modules in the software project by using the software defect prediction model. Compared with the conventional software defect prediction method, the software defect prediction model established by the method can not only predict the software defects, but also predict the severity of the defects, and has higher prediction accuracy.

Description

Software defect module severity prediction method based on ordered neural network
Technical Field
The invention belongs to the field of software quality assurance, and particularly relates to a software defect module severity degree prediction method based on an ordered neural network.
Background
The rapid development of the information technology continuously improves the complexity of the software, continuously increases the scale of the software, and is dependent on the information technology, so that the guarantee of the software quality becomes more important. Particularly in some important fields, a software system may bring about a very huge loss once it fails, and when some defects with higher severity exist in the software system, serious consequences such as software errors, functional failures, and crashes may be caused. Therefore, the detection of the module with higher severity in the software system is realized in the early stage of the software, the production and maintenance cost of the product can be reduced, and the competitiveness of an enterprise can be enhanced. Therefore, the prediction of the severity of software defect modules has led to extensive research.
Currently, for software defect module severity prediction, most researchers use a logistic regression method and a machine learning classification method to predict, and determine the severity level of a software defect according to the number of the software defects, which is equivalent to a ranking method based on a software defect severity level label, but such research does not consider the rank of the software defect severity level label itself.
It can be seen that considering the order among the software defect module severity level labels is also a problem affecting the prediction of the severity of the software defect.
Disclosure of Invention
In view of this, the present invention provides a method for predicting severity of software defect module based on ordered neural network, so as to improve accuracy of predicting severity of software project program module defect.
In order to solve the technical problem, the invention provides a software defect module severity prediction method based on an ordered neural network, which comprises the following steps:
s1, mining a version control system and a defect tracking system where a sample software project is located, marking the severity of defects of program modules in the sample software project, measuring the program modules in the sample software project by using a measuring element to obtain a sample data set, and dividing the sample data set into a training set and a verification set;
s2, constructing a neural network model according to preset weight and hyper-parameters, wherein the hyper-parameters comprise hidden node numbers, iteration times and regularization coefficients of loss functions;
s3, training the neural network model obtained in the previous step by adopting the training set to obtain an ordered neural network model;
S4, based on a verification set, adopting the ordered neural network model to predict, calculating an average absolute error, obtaining a better hyperparameter according to the average absolute error and the hyperparameter of the ordered neural network model by using a Bayesian hyperparameter optimization method, and reconstructing the neural network model according to the better hyperparameter;
s5, repeating the steps S3 and S4 until the preset repetition times are reached to obtain a neural network model with the optimal hyper-parameters, and training the neural network model with the optimal hyper-parameters by adopting the training set to obtain a software defect prediction model;
s6, predicting the severity of the defects of the program modules in the software project by using the software defect prediction model;
wherein, step S3 specifically includes the following steps:
s31, inputting values of all measuring elements of each program module in the training set into a neural network model, carrying out forward propagation to obtain output of the neural network model, and converting the output into a defect severity grade prediction value of each program module in the training set;
s32, performing back propagation training on the neural network model by using a gradient descent algorithm according to a loss function shown as the following formula to obtain an ordered neural network model;
Figure BDA0002585297060000011
Wherein y is the actual value of the defect severity grade of each program module in the training set,
Figure BDA0002585297060000012
and lambda is a regularization coefficient, and w is a weight matrix of the neural network model, for the defect severity level prediction value of each program module in the training set.
Further, in step S4, the predicting is performed by using the ordered neural network model based on the verification set, and the average absolute error is calculated, specifically:
inputting the values of all the measurement elements of each program module in the verification set into an ordered neural network model to obtain the predicted value of the defect severity grade of each program module, and calculating the average absolute error according to the actual value of the defect severity grade of each program module in the verification set, wherein the calculation formula of the average absolute error is as follows:
Figure BDA0002585297060000021
wherein MAE is the mean absolute error, m is the number of program modules contained in the verification set, yiIs the actual value of the defect severity level for the ith program module,
Figure BDA0002585297060000022
the predicted value of the defect severity level of the ith program module is within the range of 0, Q-1]And Q is the number of defect severity levels.
Further, in step S31, the converting the output into the predicted value of the severity level of the defect of each program module in the training set specifically includes: and scanning the outputs in sequence, and taking the node index with the first output value smaller than 0.5 as the predicted value of the defect severity grade of the corresponding program module.
Compared with the prior art, the method comprises the steps of firstly initializing weights and a group of hyper-parameters for constructing a neural network model; and training the neural network model by using the training set, resetting the output rule of the neural network model, adding the order of the output result, obtaining the ordered neural network model, further realizing the prediction of the severity of the defects of the program module, and having higher accuracy compared with other prediction models. The learning effect of the neural network model can be ensured by adopting a gradient descent algorithm when the neural network model is trained; obtaining the average absolute error of the ordered neural network model by using the verification set; and obtaining a better hyperparameter in a hyperparameter search space by utilizing a Bayesian hyperparameter optimization method according to the average absolute error and the hyperparameter, and reconstructing the neural network model according to the better hyperparameter. Then, training the constructed neural network model by using a training set to obtain an ordered neural network model, calculating the average absolute error of the ordered neural network model by using a verification set, then optimizing the hyperparameter of the neural network model by using a Bayesian hyperparameter optimization method to obtain a better hyperparameter, reconstructing the neural network model according to the better hyperparameter, training the constructed neural network model by using the training set to obtain the ordered neural network model, circulating the above steps until the preset number of times of repetition is reached, stopping the circulation to obtain the neural network model with the hyperparameter, and training the neural network model with the optimal hyperparameter by using the training set to obtain the optimal ordered neural network model (software defect prediction model). The invention is beneficial to selecting the ordered neural network model with the optimal hyper-parameter through the Bayesian hyper-parameter optimization method, and further improves the prediction performance of the model on the software defect module.
Drawings
FIG. 1 is a schematic flow chart of the method for predicting the severity of a software defect module based on an ordered neural network provided by the invention.
Detailed Description
For further understanding of the present invention, the method for predicting the severity of a software defect module based on an ordered neural network provided by the present invention is described in detail below with reference to the following examples, and the scope of the present invention is not limited by the following examples.
Referring to fig. 1, the present invention provides a method for predicting the severity of a software defect module based on a neural network, comprising the following steps:
s110: the program modules of the sample software project are measured by using the measurement elements, and the granularity of the program modules can be set to files or classes according to the project development language (namely, if C/C + + programming language is analyzed, the granularity of the program modules is set to files, and if Java programming language is analyzed, the granularity of the program modules is set to classes).
In the embodiment of the present invention, the program module is measured from the perspective of program size measurement, structure complexity, Halstead and CK, and specifically, as shown in table 1, the measurement element includes all rows of the module, rows containing module source code, rows containing executable source code of the module, maximum loop complexity of all nested functions in the module, sum of loop complexity of all nested functions in the module, maximum nesting level of control structure in the module, total number of different operators of the function, total number of different operands of the function, total number of vocabularies of the function, total number of operators of the function, length of the Halstead program, size of the Halstead program, difficulty of the Halstead program, level of the Halstead program, Halstead programming element, Halstead programming time, number of methods implemented in a given class, length of longest path from a given class to root in the hierarchy, The number of classes that inherit directly from a given class, the number of different non-inheriting related classes coupled to the given class, the number of methods that may be performed in response to messages received by objects of the given class, and the percentage of methods in the class calculated using the attributes for each attribute in the given class.
TABLE 1 metric element
Figure BDA0002585297060000023
Figure BDA0002585297060000031
S120: and mining a version control system and a defect tracking system where the sample software project is located, and marking the severity of the defects of the program modules in the sample software project to obtain a sample data set.
A version control system and a defect tracking system where sample software items are mined firstly extract change logs related to defect repair from the version control system by means of keywords (such as bug), and then acquire corresponding defect report information from the defect tracking system according to defect ID numbers in the change logs. The severity level of the defect is classified according to the level classification of the defect tracking system. Specifically, in the Bugzilla bug tracking system, bug program modules are divided into 7 types: the method comprises the steps of partitioning the types of the block, critic and major into severity, partitioning the normal into medium degree, and partitioning the minor, critic and enhancement into general degree); in the Jira defect tracking system, the defect program modules are divided into 5 types: the embodiment of the invention divides the block and the critic into severity, the major into medium severity and the minor into general severity). The program modules involved in the defect report information are then analyzed and labeled as corresponding defect severity according to the above-described partitioning method. Program modules for the remaining unmarked severity are all set as non-defective program modules.
In an experiment, a sample data set of four versions, namely 1.4, 1.5, 1.6 and 1.7, of open source software Apache Ant is constructed, a measurement element shown in a table 1 is adopted to measure each program module, then all the program modules are marked as four types, namely a serious defect type, a medium defect type, a general defect type and a non-defect type through the marking process, the types are respectively shown as 4, 3, 2 and 1, a sample data set is generated, and part of the data set is shown in a table 2;
table 2 partial data set presentation
Figure BDA0002585297060000032
Figure BDA0002585297060000041
S130: the sample data set is divided into a training set and a verification set.
S140: constructing a neural network model according to preset weight and hyper-parameters, wherein the hyper-parameters comprise the number of hidden nodes, iteration times and regularization coefficients of a loss function; training the neural network model by adopting a training set to obtain an ordered neural network model;
specifically, in the embodiment of the present invention, the weights, the number of hidden nodes, the number of iterations, and the regularization coefficients of the loss function between layers in the neural network model are initialized, and the neural network model is constructed. In the embodiment of the invention, the neural network model structure is as follows:
(1) input layer
The number of input layers is 1, and the input layers are provided with 23 neuron nodes, wherein 23 is the number of the measurement elements.
(2) Hidden layer
The number of hidden layer layers is 1, and the number of neuron nodes is self-defined to be 50 (the number of hidden layer nodes).
(3) Output layer
The number of output layers is 1, and the number of nodes is the number of defect severity levels of the program module, in the embodiment of the invention, the number of defect severity levels of the program module is 4, that is, the number of nodes is 4. In the embodiment of the invention, the activation function of each layer of the neural network model adopts a sigmoid function. The number of iterations is preset to 200.
The number of hidden layer nodes, the iteration times and the regularization coefficient of a loss function of the neural network model belong to parameters to be optimized by a Bayes super-parametric optimization method, except for the first random initialization, the other parameters are automatically selected and assigned by the Bayes super-parametric optimization method, and the weights among layers in the neural network model are randomly generated when the neural network model is trained each time.
And inputting the values of all the measurement elements of each program module in the training set into the neural network model, carrying out forward propagation to obtain the output of the neural network model, and converting the output result into the predicted value of the defect severity grade of each program module. And the conversion rule is to scan each output node in sequence and take the node index with the first output value smaller than 0.5 as the predicted value of the defect severity grade of the corresponding program module.
In this embodiment, the severity level of the defect of the program module is set to 4, so that 4 nodes exist in the output layer of the neural network model, and all the measurement elements of a certain program module in the training set are input into the neural network model and are propagated in the forward direction to obtain the output of the neural network model. If the output of the neural network model is [0.9, 0.7, 0.3, 0.2], 0.3 is the first node less than 0.5, and the index value is 3, the predicted value of the defect severity level of the sample is 3; if the output of the neural network model is [0.6, 0.3, 0.7, 0.9], and then 0.3 is the first node less than 0.5, and the index value is 2, then the predicted value of the defect severity level for this sample is 2.
Determining a loss function loss according to the predicted value of the severity grade of the defect of each program module in the training set and the actual value of the severity grade of the defect of each program module in the training set, wherein
Figure BDA0002585297060000042
y is the actual value of the defect severity level of each program module in the training set,
Figure BDA0002585297060000043
and lambda is a regularization coefficient, and w is a weight matrix of the neural network model, for the defect severity level prediction value of each program module in the training set.
And carrying out back propagation training on the neural network model by using a gradient descent algorithm according to the loss function loss to obtain the ordered neural network model.
S150: inputting the values of all the measurement elements of each program module in the verification set into the ordered neural network model to obtain the predicted value of the defect severity grade of each program module, and calculating the average absolute error according to the actual value of the defect severity grade of each program module in the verification set, wherein the calculation formula of the average absolute error is as follows:
Figure BDA0002585297060000044
wherein MAE is the mean absolute error, m is the number of program modules contained in the verification set, yiIs the actual value of the defect severity level for the ith program module,
Figure BDA0002585297060000045
the predicted value of the defect severity level of the ith program module is within the range of 0, Q-1]Q is the number of defect severity levels; in the embodiment of the present invention, the range of variation of MAE is [0, 3 ]]。
S160: based on the average absolute error and the hyperparameters of the ordered neural network model, obtaining a better hyperparameter by utilizing Bayesian hyperparametric optimization, and reconstructing the neural network model according to the better hyperparameter;
the parameters required by the Bayesian super-parameter optimization method are as follows:
a. an objective function: mean absolute error of the ordered neural network model;
b. hyper-parametric search space: the experiment comprises iteration times, hidden node number and regularization coefficient; the number of self-defined iterations is [200,500] and 25 is taken as an interval value, the number of hidden nodes is [5,50] and 5 is taken as an interval value, and the regularization coefficient is [0,1] and is taken randomly;
c. And (3) an optimization algorithm: and a Bayesian optimization algorithm based on a Gaussian process and a regression tree. The algorithm is a built-in algorithm of a Bayesian super-parameter optimization method.
Calculating the average absolute error of the ordered neural network model by using the verification set, then optimizing the hyper-parameters of the neural network model by using a Bayesian hyper-parameter optimization method to obtain a better hyper-parameter, reconstructing the neural network model according to the better hyper-parameter, training the reconstructed neural network model by using the training set to obtain the ordered neural network model, repeating the steps until the preset number of times is reached, stopping the circulation to obtain the neural network model with the optimal hyper-parameter, and training the neural network model with the optimal hyper-parameter by using the training set to obtain a software defect prediction model, namely the optimal ordered neural network model. In the present embodiment, the number of repetitions is set to 100.
S170: and processing the program modules of the software project to be predicted by adopting a software defect prediction model to obtain the severity grade of the defects of each program module in the software project to be predicted.
In the embodiment, an experiment is carried out by adopting four versions of an Apache Ant project, a cross-version experiment scene is adopted, namely a data set of a previous version is used for training a software defect prediction model, and the built model is evaluated by using a data set of a current version. The method specifically comprises the following steps: 1. and (3) constructing a software defect prediction model by using the data set of Apache Ant 1.4 version, and performing performance evaluation on the constructed model by using the data set of Apache Ant 1.5 version. 2. And (3) constructing a software defect prediction model by using the data set of Apache Ant 1.5 version, and performing performance evaluation on the constructed model by using the data set of Apache Ant 1.6 version. 3. The software defect prediction model is built by using the data set of Apache Ant 1.6 version, and the performance of the built model is evaluated by using the data set of Apache Ant1.7 version.
In order to detect the experimental results of the invention, two multi-classification methods of decision trees and K neighbors are adopted for comparison, and the specific results are shown in Table 3.
TABLE 3 mean absolute error of different versions of Apache Ant
Figure BDA0002585297060000051
While there have been shown and described what are at present considered the fundamental principles and essential features of the invention and its advantages, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing exemplary embodiments, but is capable of other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (3)

1. A software defect module severity prediction method based on an ordered neural network is characterized by comprising the following steps:
s1, mining a version control system and a defect tracking system where a sample software project is located, marking the severity of defects of program modules in the sample software project, measuring the program modules in the sample software project by using a measuring element to obtain a sample data set, and dividing the sample data set into a training set and a verification set;
s2, constructing a neural network model according to preset weight and hyper-parameters, wherein the hyper-parameters comprise hidden node numbers, iteration times and regularization coefficients of loss functions;
s3, training the neural network model obtained in the previous step by adopting the training set to obtain an ordered neural network model;
s4, based on a verification set, adopting the ordered neural network model to predict, calculating an average absolute error, obtaining a better hyperparameter according to the average absolute error and the hyperparameter of the ordered neural network model by using a Bayesian hyperparameter optimization method, and reconstructing the neural network model according to the better hyperparameter;
s5, repeating the steps S3 and S4 until the preset repetition times are reached to obtain a neural network model with the optimal hyper-parameters, and training the neural network model with the optimal hyper-parameters by adopting the training set to obtain a software defect prediction model;
S6, predicting the severity of the defects of the program modules in the software project by using the software defect prediction model;
wherein, step S3 specifically includes the following steps:
s31, inputting values of all measuring elements of each program module in the training set into a neural network model, carrying out forward propagation to obtain output of the neural network model, and converting the output into a defect severity grade prediction value of each program module in the training set;
s32, performing back propagation training on the neural network model by using a gradient descent algorithm according to a loss function shown as the following formula to obtain an ordered neural network model;
Figure FDA0002585297050000011
wherein g is the actual value of the defect severity grade of each program module in the training set,
Figure FDA0002585297050000013
and lambda is a regularization coefficient, and w is a weight matrix of the neural network model, for the defect severity level prediction value of each program module in the training set.
2. The method for predicting the severity of a software defect module according to claim 1, wherein in step S4, the ordered neural network model is used for prediction based on the validation set, and the average absolute error is calculated, specifically:
inputting the values of all the measurement elements of each program module in the verification set into an ordered neural network model to obtain the predicted value of the defect severity grade of each program module, and calculating the average absolute error according to the actual value of the defect severity grade of each program module in the verification set, wherein the calculation formula of the average absolute error is as follows:
Figure FDA0002585297050000012
Wherein MAE is the mean absolute error, m is the number of program modules contained in the verification set, yiFor the ith program moduleThe actual value of the defect severity level of the block,
Figure FDA0002585297050000014
the predicted value of the defect severity level of the ith program module is within the range of 0, Q-1]And Q is the number of defect severity levels.
3. The method of claim 1, wherein in step S31, the converting the output into the predicted value of the severity level of the defect for each program module in the training set includes: and scanning the outputs in sequence, and taking the node index with the first output value smaller than 0.5 as the predicted value of the defect severity grade of the corresponding program module.
CN202010679603.4A 2020-07-15 2020-07-15 Software defect module severity prediction method based on ordered neural network Active CN111858328B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010679603.4A CN111858328B (en) 2020-07-15 2020-07-15 Software defect module severity prediction method based on ordered neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010679603.4A CN111858328B (en) 2020-07-15 2020-07-15 Software defect module severity prediction method based on ordered neural network

Publications (2)

Publication Number Publication Date
CN111858328A true CN111858328A (en) 2020-10-30
CN111858328B CN111858328B (en) 2021-11-12

Family

ID=72983501

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010679603.4A Active CN111858328B (en) 2020-07-15 2020-07-15 Software defect module severity prediction method based on ordered neural network

Country Status (1)

Country Link
CN (1) CN111858328B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113378907A (en) * 2021-06-04 2021-09-10 南京大学 Automatic software traceability recovery method for enhancing data preprocessing process
CN114528221A (en) * 2022-02-24 2022-05-24 北京航空航天大学 Software defect prediction method based on heterogeneous graph neural network
CN115269377A (en) * 2022-06-23 2022-11-01 南通大学 Cross-project software defect prediction method based on optimization instance selection

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105608004A (en) * 2015-12-17 2016-05-25 云南大学 CS-ANN-based software failure prediction method
EP3509019A1 (en) * 2018-01-05 2019-07-10 Accenture Global Solutions Limited Software product development defect and issue prediction and diagnosis
CN110162478A (en) * 2019-05-28 2019-08-23 南京大学 A kind of defect code path orientation method based on defect report
CN111368885A (en) * 2020-02-24 2020-07-03 大连理工大学 Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105608004A (en) * 2015-12-17 2016-05-25 云南大学 CS-ANN-based software failure prediction method
EP3509019A1 (en) * 2018-01-05 2019-07-10 Accenture Global Solutions Limited Software product development defect and issue prediction and diagnosis
CN110162478A (en) * 2019-05-28 2019-08-23 南京大学 A kind of defect code path orientation method based on defect report
CN111368885A (en) * 2020-02-24 2020-07-03 大连理工大学 Aero-engine gas circuit fault diagnosis method based on deep learning and information fusion

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘树毅等: ""基于域自适应神经网络的跨项目软件缺陷预测"", 《计算机与数字工程》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113378907A (en) * 2021-06-04 2021-09-10 南京大学 Automatic software traceability recovery method for enhancing data preprocessing process
CN113378907B (en) * 2021-06-04 2024-01-09 南京大学 Automated software traceability recovery method for enhancing data preprocessing process
CN114528221A (en) * 2022-02-24 2022-05-24 北京航空航天大学 Software defect prediction method based on heterogeneous graph neural network
CN115269377A (en) * 2022-06-23 2022-11-01 南通大学 Cross-project software defect prediction method based on optimization instance selection

Also Published As

Publication number Publication date
CN111858328B (en) 2021-11-12

Similar Documents

Publication Publication Date Title
CN111858328B (en) Software defect module severity prediction method based on ordered neural network
US20210342699A1 (en) Cooperative execution of a genetic algorithm with an efficient training algorithm for data-driven model creation
US11687786B2 (en) Pre-processing for data-driven model creation
Xia et al. Collective personalized change classification with multiobjective search
US20190130277A1 (en) Ensembling of neural network models
Cano et al. Evolving rule-based classifiers with genetic programming on GPUs for drifting data streams
KR102531645B1 (en) Computer program for performance testing of models
Edali et al. Exploring the behavior space of agent-based simulation models using random forest metamodels and sequential sampling
US11460982B1 (en) Number embedding application system
US20210374544A1 (en) Leveraging lagging gradients in machine-learning model training
Cruz et al. Detecting bad smells with machine learning algorithms: An empirical study
KR20210107491A (en) Method for generating anoamlous data
Zheng et al. A self-adaptive temporal-spatial self-training algorithm for semisupervised fault diagnosis of industrial processes
Shafiq et al. Retracted: Scientific programming using optimized machine learning techniques for software fault prediction to improve software quality
JP2023510922A (en) Optimizing High-Cost Functions for Complex Multidimensional Constraints
Abdelkarim et al. Tcp-net: Test case prioritization using end-to-end deep neural networks
Amorim et al. A new word embedding approach to evaluate potential fixes for automated program repair
CN116739100A (en) Vulnerability detection method of quantum neural network and automatic driving vulnerability detection method
Fazakis et al. An active learning ensemble method for regression tasks
CN112380132A (en) Countermeasure verification method and device based on class unbalance of defect data set of aerospace software
US20230419086A1 (en) System and method for processing material properties of structural materials
Mulla et al. The Use of Clustering and Classification Methods in Machine Learning and Comparison of Some Algorithms of the Methods
Anjali et al. Moth Flame Optimization Based FCNN for Prediction of Bugs in Software.
CN112416789B (en) Process metric element evaluation method for evolution software
Agnani et al. Predicting E-commerce Sales & Inventory Management using Machine Learning

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