Empirical analysis of network measures for effort-aware fault-proneness prediction

ContextRecently, network measures have been proposed to predict fault-prone modules. Leveraging the dependency relationships between software entities, network measures describe the structural features of software systems. However, there is no consensus about their effectiveness for fault-proneness prediction. Specifically, the predictive ability of network measures in effort-aware context has not been addressed.ObjectiveWe aim to provide a comprehensive evaluation on the predictive effectiveness of network measures with the effort needed to inspect the code taken into consideration.MethodWe first constructed software source code networks of 11 open-source projects by extracting the data and call dependencies between modules. We then employed univariate logistic regression to investigate how each single network measure was correlated with fault-proneness. Finally, we built multivariate prediction models to examine the usefulness of network measures under three prediction settings: cross-validation, across-release, and inter-project predictions. In particular, we used the effort-aware performance indicators to compare their predictive ability against the commonly used code metrics in both ranking and classification scenarios.ResultsBased on the 11 open-source software systems, our results show that: (1) most network measures are significantly positively related to fault-proneness; (2) the performance of network measures varies under different prediction settings; (3) network measures have inconsistent effects on various projects.ConclusionNetwork measures are of practical value in the context of effort-aware fault-proneness prediction, but researchers and practitioners should be careful of choosing whether and when to use network measures in practice.     

Empirical Analysis of Object-Oriented Design Metrics for Predicting High and Low Severity Faults

In the last decade, empirical studies on object-oriented design metrics have shown some of them to be useful for predicting the fault-proneness of classes in object-oriented software systems. This research did not, however, distinguish among faults according to the severity of impact. It would be valuable to know how object-oriented design metrics and class fault-proneness are related when fault severity is taken into account. In this paper, we use logistic regression and machine learning methods to empirically investigate the usefulness of object-oriented design metrics, specifically, a subset of the Chidamber and Kemerer suite, in predicting fault-proneness when taking fault severity into account. Our results, based on a public domain NASA data set, indicate that 1) most of these design metrics are statistically related to fault-proneness of classes across fault severity, and 2) the prediction capabilities of the investigated metrics greatly depend on the severity of faults. More specifically, these design metrics are able to predict low severity faults in fault-prone classes better than high severity faults in fault-prone classes     

Prediction approach of software fault-proneness based on hybrid artificial neural network and quantum particle swarm optimization

The identification of a module's fault-proneness is very important for minimizing cost and improving the effectiveness of the software development process. How to obtain the correlation between software metrics and module's fault-proneness has been the focus of much research. This paper presents the application of hybrid artificial neural network (ANN) and Quantum Particle Swarm Optimization (QPSO) in software fault-proneness prediction. ANN is used for classifying software modules into fault-proneness or non fault-proneness categories, and QPSO is applied for reducing dimensionality. The experiment results show that the proposed prediction approach can establish the correlation between software metrics and modules’ fault-proneness, and is very simple because its implementation requires neither extra cost nor expert's knowledge. Proposed prediction approach can provide the potential software modules with fault-proneness to software developers, so developers only need to focus on these software modules, which may minimize effort and cost of software maintenance.     

基于支持向量机的面向对象软件易发性故障预测

本文考虑软件故障严重程度,并采用C＆K面向对象度量集,以支持向量机分析方法为数学工具,建立一种基于面向对象软件易发性故障预测模型。实验结果表明,与基于朴素贝叶斯的预测模型、随机预测模型和NNge预测模型相比,本文提出的预测模型对于高严重程度故障、低严重程度故障以及未划分故障严重程度的情形均获得较好的预测效果。     

An empirical analysis of package-modularization metrics: Implications for software fault-proneness

ContextIn a large object-oriented software system, packages play the role of modules which group related classes together to provide well-identified services to the rest of the system. In this context, it is widely believed that modularization has a large influence on the quality of packages. Recently, Sarkar, Kak, and Rama proposed a set of new metrics to characterize the modularization quality of packages from important perspectives such as inter-module call traffic, state access violations, fragile base-class design, programming to interface, and plugin pollution. These package-modularization metrics are quite different from traditional package-level metrics, which measure software quality mainly from size, extensibility, responsibility, independence, abstractness, and instability perspectives. As such, it is expected that these package-modularization metrics should be useful predictors for fault-proneness. However, little is currently known on their actual usefulness for fault-proneness prediction, especially compared with traditional package-level metrics.ObjectiveIn this paper, we examine the role of these new package-modularization metrics for determining software fault-proneness in object-oriented systems.MethodWe first use principal component analysis to analyze whether these new package-modularization metrics capture additional information compared with traditional package-level metrics. Second, we employ univariate prediction models to investigate how these new package-modularization metrics are related to fault-proneness. Finally, we build multivariate prediction models to examine the ability of these new package-modularization metrics for predicting fault-prone packages.ResultsOur results, based on six open-source object-oriented software systems, show that: (1) these new package-modularization metrics provide new and complementary views of software complexity compared with traditional package-level metrics; (2) most of these new package-modularization metrics have a significant association with fault-proneness in an expected direction; and (3) these new package-modularization metrics can substantially improve the effectiveness of fault-proneness prediction when used with traditional package-level metrics together.ConclusionsThe package-modularization metrics proposed by Sarkar, Kak, and Rama are useful for practitioners to develop quality software systems.     

On the ability of complexity metrics to predict fault-prone classes in object-oriented systems

Many studies use logistic regression models to investigate the ability of complexity metrics to predict fault-prone classes. However, it is not uncommon to see the inappropriate use of performance indictors such as odds ratio in previous studies. In particular, a recent study by Olague et al. uses the odds ratio associated with one unit increase in a metric to compare the relative magnitude of the associations between individual metrics and fault-proneness. In addition, the percents of concordant, discordant, and tied pairs are used to evaluate the predictive effectiveness of a univariate logistic regression model. Their results suggest that lesser known complexity metrics such as standard deviation method complexity (SDMC) and average method complexity (AMC) are better predictors than the two commonly used metrics: lines of code (LOC) and weighted method McCabe complexity (WMC). In this paper, however, we show that (1) the odds ratio associated with one standard deviation increase, rather than one unit increase, in a metric should be used to compare the relative magnitudes of the effects of individual metrics on fault-proneness. Otherwise, misleading results may be obtained; and that (2) the connection of the percents of concordant, discordant, and tied pairs with the predictive effectiveness of a univariate logistic regression model is false, as they indeed do not depend on the model. Furthermore, we use the data collected from three versions of Eclipse to re-examine the ability of complexity metrics to predict fault-proneness. Our experimental results reveal that: (1) many metrics exhibit moderate or almost moderate ability in discriminating between fault-prone and not fault-prone classes; (2) LOC and WMC are indeed better fault-proneness predictors than SDMC and AMC; and (3) the explanatory power of other complexity metrics in addition to LOC is limited.     

Deriving thresholds of software metrics to predict faults on open source software: Replicated case studies

Object-oriented metrics aim to exhibit the quality of source code and give insight to it quantitatively. Each metric assesses the code from a different aspect. There is a relationship between the quality level and the risk level of source code. The objective of this paper is to empirically examine whether or not there are effective threshold values for source code metrics. It is targeted to derive generalized thresholds that can be used in different software systems. The relationship between metric thresholds and fault-proneness was investigated empirically in this study by using ten open-source software systems. Three types of fault-proneness were defined for the software modules: non-fault-prone, more-than-one-fault-prone, and more-than-three-fault-prone. Two independent case studies were carried out to derive two different threshold values. A single set was created by merging ten datasets and was used as training data by the model. The learner model was created using logistic regression and the Bender method. Results revealed that some metrics have threshold effects. Seven metrics gave satisfactory results in the first case study. In the second case study, eleven metrics gave satisfactory results. This study makes contributions primarily for use by software developers and testers. Software developers can see classes or modules that require revising; this, consequently, contributes to an increment in quality for these modules and a decrement in their risk level. Testers can identify modules that need more testing effort and can prioritize modules according to their risk levels.     

A systematic and comprehensive investigation of methods to build and evaluate fault prediction models

This paper describes a study performed in an industrial setting that attempts to build predictive models to identify parts of a Java system with a high fault probability. The system under consideration is constantly evolving as several releases a year are shipped to customers. Developers usually have limited resources for their testing and would like to devote extra resources to faulty system parts. The main research focus of this paper is to systematically assess three aspects on how to build and evaluate fault-proneness models in the context of this large Java legacy system development project: (1) compare many data mining and machine learning techniques to build fault-proneness models, (2) assess the impact of using different metric sets such as source code structural measures and change/fault history (process measures), and (3) compare several alternative ways of assessing the performance of the models, in terms of (i) confusion matrix criteria such as accuracy and precision/recall, (ii) ranking ability, using the receiver operating characteristic area (ROC), and (iii) our proposed cost-effectiveness measure (CE).The results of the study indicate that the choice of fault-proneness modeling technique has limited impact on the resulting classification accuracy or cost-effectiveness. There is however large differences between the individual metric sets in terms of cost-effectiveness, and although the process measures are among the most expensive ones to collect, including them as candidate measures significantly improves the prediction models compared with models that only include structural measures and/or their deltas between releases - both in terms of ROC area and in terms of CE. Further, we observe that what is considered the best model is highly dependent on the criteria that are used to evaluate and compare the models. And the regular confusion matrix criteria, although popular, are not clearly related to the problem at hand, namely the cost-effectiveness of using fault-proneness prediction models to focus verification efforts to deliver software with less faults at less cost.     

Applying machine learning to software fault-proneness prediction

The importance of software testing to quality assurance cannot be overemphasized. The estimation of a module’s fault-proneness is important for minimizing cost and improving the effectiveness of the software testing process. Unfortunately, no general technique for estimating software fault-proneness is available. The observed correlation between some software metrics and fault-proneness has resulted in a variety of predictive models based on multiple metrics. Much work has concentrated on how to select the software metrics that are most likely to indicate fault-proneness. In this paper, we propose the use of machine learning for this purpose. Specifically, given historical data on software metric values and number of reported errors, an Artificial Neural Network (ANN) is trained. Then, in order to determine the importance of each software metric in predicting fault-proneness, a sensitivity analysis is performed on the trained ANN. The software metrics that are deemed to be the most critical are then used as the basis of an ANN-based predictive model of a continuous measure of fault-proneness. We also view fault-proneness prediction as a binary classification task (i.e., a module can either contain errors or be error-free) and use Support Vector Machines (SVM) as a state-of-the-art classification method. We perform a comparative experimental study of the effectiveness of ANNs and SVMs on a data set obtained from NASA’s Metrics Data Program data repository.     

Empirical validation of object-oriented metrics on open source software for fault prediction

Open source software systems are becoming increasingly important these days. Many companies are investing in open source projects and lots of them are also using such software in their own work. But, because open source software is often developed with a different management style than the industrial ones, the quality and reliability of the code needs to be studied. Hence, the characteristics of the source code of these projects need to be measured to obtain more information about it. This paper describes how we calculated the object-oriented metrics given by Chidamber and Kemerer to illustrate how fault-proneness detection of the source code of the open source Web and e-mail suite called Mozilla can be carried out. We checked the values obtained against the number of bugs found in its bug database - called Bugzilla - using regression and machine learning methods to validate the usefulness of these metrics for fault-proneness prediction. We also compared the metrics of several versions of Mozilla to see how the predicted fault-proneness of the software system changed during its development cycle.     

Estimating software readiness using predictive models

In this study, defect tracking is used as a proxy method to predict software readiness. The number of remaining defects in an application under development is one of the most important factors that allow one to decide if a piece of software is ready to be released. By comparing predicted number of faults and number of faults discovered in testing, software manager can decide whether the software is likely ready to be released or not.The predictive model developed in this research can predict: (i) the number of faults (defects) likely to exist, (ii) the estimated number of code changes required to correct a fault and (iii) the estimated amount of time (in minutes) needed to make the changes in respective classes of the application. The model uses product metrics as independent variables to do predictions. These metrics are selected depending on the nature of source code with regards to architecture layers, types of faults and contribution factors of these metrics. The use of neural network model with genetic training strategy is introduced to improve prediction results for estimating software readiness in this study. This genetic-net combines a genetic algorithm with a statistical estimator to produce a model which also shows the usefulness of inputs.The model is divided into three parts: (1) prediction model for presentation logic tier (2) prediction model for business tier and (3) prediction model for data access tier. Existing object-oriented metrics and complexity software metrics are used in the business tier prediction model. New sets of metrics have been proposed for the presentation logic tier and data access tier. These metrics are validated using data extracted from real world applications. The trained models can be used as tools to assist software mangers in making software release decisions.     

Fault Prediction Modeling for Software Quality Estimation: Comparing Commonly Used Techniques

High-assurance and complex mission-critical software systems are heavily dependent on reliability of their underlying software applications. An early software fault prediction is a proven technique in achieving high software reliability. Prediction models based on software metrics can predict number of faults in software modules. Timely predictions of such models can be used to direct cost-effective quality enhancement efforts to modules that are likely to have a high number of faults. We evaluate the predictive performance of six commonly used fault prediction techniques: CART-LS (least squares), CART-LAD (least absolute deviation), S-PLUS, multiple linear regression, artificial neural networks, and case-based reasoning. The case study consists of software metrics collected over four releases of a very large telecommunications system. Performance metrics, average absolute and average relative errors, are utilized to gauge the accuracy of different prediction models. Models were built using both, original software metrics (RAW) and their principle components (PCA). Two-way ANOVA randomized-complete block design models with two blocking variables are designed with average absolute and average relative errors as response variables. System release and the model type (RAW or PCA) form the blocking variables and the prediction technique is treated as a factor. Using multiple-pairwise comparisons, the performance order of prediction models is determined. We observe that for both average absolute and average relative errors, the CART-LAD model performs the best while the S-PLUS model is ranked sixth.     

Application of neural networks for predicting program faults

Accurately predicting the number of faults in program modules is a major problem in the quality control of large software development efforts. Some software complexity metrics are closely related to the distribution of faults across program modules. Using these relationships, software engineers develop models that provide early estimates of quality metrics that do not become available until late in the development cycle. By considering these early estimates, software engineers can take actions to avoid or prepare for emerging quality problems. Most often, the predictive models are based upon multiple regression analysis. However, measures of software quality and complexity exhibit systematic departures from the assumptions of these analyses. With extreme violations of these assumptions, multiple regression models become unstable and lose most of their predictive quality. Since neural network models carry no data assumptions, these models could be more appropriate than regression models for modeling software faults. In this paper, we explore a neural network methodology for developing models that predict the number of faults in program modules. We apply this methodology to develop neural network models based upon data collected during the development of two commercial software systems. After developing neural network models, we apply multiple linear regression methods to develop regression models on the same data. For the data sets considered, the neural network methodology produced better predictive models in terms of both quality of fit and predictive quality.     

Empirical validation of object-oriented metrics for predicting fault proneness models

Empirical validation of software metrics used to predict software quality attributes is important to ensure their practical relevance in software organizations. The aim of this work is to find the relation of object-oriented (OO) metrics with fault proneness at different severity levels of faults. For this purpose, different prediction models have been developed using regression and machine learning methods. We evaluate and compare the performance of these methods to find which method performs better at different severity levels of faults and empirically validate OO metrics given by Chidamber and Kemerer. The results of the empirical study are based on public domain NASA data set. The performance of the predicted models was evaluated using Receiver Operating Characteristic (ROC) analysis. The results show that the area under the curve (measured from the ROC analysis) of models predicted using high severity faults is low as compared with the area under the curve of the model predicted with respect to medium and low severity faults. However, the number of faults in the classes correctly classified by predicted models with respect to high severity faults is not low. This study also shows that the performance of machine learning methods is better than logistic regression method with respect to all the severities of faults. Based on the results, it is reasonable to claim that models targeted at different severity levels of faults could help for planning and executing testing by focusing resources on fault-prone parts of the design and code that are likely to cause serious failures.     

Modeling change requests due to faults in a large-scale telecommunication system

It is widely known that a small number of modules in any system are likely to contain the majority of faults. Early identification and consequent attention to such modules may mitigate or prevent many defects. The objective of this study is to use product metrics to build a prediction model of the number of change requests (CRs) that are likely to occur in individual modules during testing. The study first empirically validates eight product metrics, while considering the confounding effects of code size (lines of code). Next, a prediction model of CR outcomes is developed with the validated metrics by utilizing a negative binomial regression that allows over-dispersion. In total, 816 modules written in the Chill programming language were analyzed in a large-scale telecommunication system. There is a positive association between the number of CRs and four product metrics (number of unique operators, unique operands, signals, and library calls) after considering the confounding effect of code size. A prediction model that includes only code size and the number of unique operands provides the best empirical fit.     

Predicting fault-prone software modules in telephone switches

An empirical study was carried out at Ericsson Telecom AB to investigate the relationship between several design metrics and the number of function test failure reports associated with software modules. A tool, ERIMET, was developed to analyze the design documents automatically. Preliminary results from the study of 130 modules showed that: based on fault and design data one can satisfactorily build, before coding has started, a prediction model for identifying the most fault-prone modules. The data analyzed show that 20 percent of the most fault-prone modules account for 60 percent of all faults. The prediction model built in this paper would have identified 20 percent of the modules accounting for 47 percent of all faults. At least four design measures can alternatively be used as predictors with equivalent performance. The size (with respect to the number of lines of code) used in a previous prediction model was not significantly better than these four measures. The Alberg diagram introduced in this paper offers a way of assessing a predictor based on historical data, which is a valuable complement to linear regression when prediction data is ordinal. Applying the method described in this paper makes it possible to use measures at the design phase to predict the most fault-prone modules     

Using structural and textual information to capture feature coupling in object-oriented software

Previous studies have demonstrated the relationship between coupling and external software quality attributes, such as fault-proneness, and the application of coupling to software maintenance tasks, such as impact analysis. These previous studies concentrate on class coupling. However, there is a growing focus on the study of features in software, and features are often implemented across multiple classes, meaning class-level coupling measures are not applicable. We ask the pertinent question, “Is measuring coupling at the feature-level also useful?” We define new feature coupling metrics based on structural and textual source code information and extend the unified framework for coupling measurement to include these new metrics. We also conduct three extensive case studies to evaluate these new metrics and answer this research question. The first study examines the relationship between feature coupling and fault-proneness, the second assesses feature coupling in the context of impact analysis, and the third study surveys developers to determine if the metrics align with what they consider to be coupled features. All three studies provide evidence that feature coupling metrics are indeed useful new measures that capture coupling at a higher level of abstraction than classes and can be useful for finding bugs, guiding testing effort, and assessing change impact.     

Data Transformation in Cross-project Defect Prediction

Software metrics rarely follow a normal distribution. Therefore, software metrics are usually transformed prior to building a defect prediction model. To the best of our knowledge, the impact that the transformation has on cross-project defect prediction models has not been thoroughly explored. A cross-project model is built from one project and applied on another project. In this study, we investigate if cross-project defect prediction is affected by applying different transformations (i.e., log and rank transformations, as well as the Box-Cox transformation). The Box-Cox transformation subsumes log and other power transformations (e.g., square root), but has not been studied in the defect prediction literature. We propose an approach, namely Multiple Transformations (MT), to utilize multiple transformations for cross-project defect prediction. We further propose an enhanced approach MT+ to use the parameter of the Box-Cox transformation to determine the most appropriate training project for each target project. Our experiments are conducted upon three publicly available data sets (i.e., AEEEM, ReLink, and PROMISE). Comparing to the random forest model built solely using the log transformation, our MT+ approach improves the F-measure by 7, 59 and 43% for the three data sets, respectively. As a summary, our major contributions are three-fold: 1) conduct an empirical study on the impact that data transformation has on cross-project defect prediction models; 2) propose an approach to utilize the various information retained by applying different transformation methods; and 3) propose an unsupervised approach to select the most appropriate training project for each target project.     

A symbolic fault-prediction model based on multiobjective particle swarm optimization

In the literature the fault-proneness of classes or methods has been used to devise strategies for reducing testing costs and efforts. In general, fault-proneness is predicted through a set of design metrics and, most recently, by using Machine Learning (ML) techniques. However, some ML techniques cannot deal with unbalanced data, characteristic very common of the fault datasets and, their produced results are not easily interpreted by most programmers and testers. Considering these facts, this paper introduces a novel fault-prediction approach based on Multiobjective Particle Swarm Optimization (MOPSO). Exploring Pareto dominance concepts, the approach generates a model composed by rules with specific properties. These rules can be used as an unordered classifier, and because of this, they are more intuitive and comprehensible. Two experiments were accomplished, considering, respectively, fault-proneness of classes and methods. The results show interesting relationships between the studied metrics and fault prediction. In addition to this, the performance of the introduced MOPSO approach is compared with other ML algorithms by using several measures including the area under the ROC curve, which is a relevant criterion to deal with unbalanced data.     

Thresholds based outlier detection approach for mining class outliers: An empirical case study on software measurement datasets

Predicting the fault-proneness labels of software program modules is an emerging software quality assurance activity and the quality of datasets collected from previous software version affects the performance of fault prediction models. In this paper, we propose an outlier detection approach using metrics thresholds and class labels to identify class outliers. We evaluate our approach on public NASA datasets from PROMISE repository. Experiments reveal that this novel outlier detection method improves the performance of robust software fault prediction models based on Naive Bayes and Random Forests machine learning algorithms.