Cross-project defect prediction using a connectivity-based unsupervised classifier

Zhang, Feng; Zheng, Quan; Zou, Ying; Hassan, Ahmed E.

doi:10.1145/2884781.2884839

“…To construct a training target data, we randomly select 10 % of the instances in a target project, and we mitigate this potential bias with 20 runs of the experiment. (4)Evaluation bias. We use 2 comprehensive measures g‐measure and AUC , both of them have been widely used to evaluate the effectiveness of fault prediction studies . Measuring prediction performance of other measures is left for future work.…”

Section: Discussionmentioning

confidence: 99%

Heterogeneous fault prediction with cost‐sensitive domain adaptation

Li

¹

,

Jing

²

,

Zhu

³

2018

Software Testing Verif & Rel

View full text Add to dashboard Cite

Summary In the early phases of software testing, projects may have only limited historical defect data. Learning prediction model with such insufficient training data will limit the efficacy of learned predictor. In practice, there are usually many publicly available fault prediction datasets. Recently, heterogeneous fault prediction (HFP) has been proposed. However, existing HFP models do not investigate how to use mixed project data to predict target. Furthermore, defect data are often imbalanced. The imbalanced data distribution of source usually leads to serious misclassification of fault‐prone instances, which will degrade the predictor's performance. Existing HFP methods do not consider the class imbalance problem in the training stages. In this paper, we propose a novel Cost‐sensitive Label and Structure‐consistent Unilateral Projection (CLSUP) approach for HFP. CLSUP can not only make better use of the within‐project and cross‐project data but also alleviate the class imbalance problem by setting different misclassification costs for fault‐prone and non–fault‐prone instances. Extensive experiments on 30 projects demonstrate the effectiveness of CLSUP.

show abstract

“…The third mainstream way is to apply unsupervised classifier that does not require any training data to perform CCDP (e.g., [29][30]), therefore the distribution gap between the training project data and the target project data is no longer an issue. For instance, Zhang et.al [30] proposed to apply a connectivity-based unsupervised classifier that is based on Agglomerative clustering to perform CPDP.…”

Section: B Cross-company Defect Predictionmentioning

confidence: 99%

“…For instance, Zhang et.al [30] proposed to apply a connectivity-based unsupervised classifier that is based on Agglomerative clustering to perform CPDP.…”

Section: B Cross-company Defect Predictionmentioning

confidence: 99%

A Data Filtering Method Based on Agglomerative Clustering

Yu¹,

Peipei²,

Zhang³

et al. 2017

International Conferences on Software Engineering and Knowledge Engineering

View full text Add to dashboard Cite

Abstract-Cross-company defect prediction (CCDP) is a practical way that trains a prediction model by exploiting one or multiple projects of a source company and then applies the model to target company. Unfortunately, larger irrelevant crosscompany (CC) data usually makes it difficult to build a crosscompany defect prediction model with high performance. To address such issues, this paper proposes a data filtering method based on Agglomerative Clustering (DFAC) for cross-company defect prediction. First, DFAC combines within-company instances and cross-company instances and uses Agglomerative clustering algorithms to group these instances. Second, DFAC selects sub-clusters which consist at least one WC instance, and collects the CC instances in the selected sub-clusters into a new CC data. Compared with existing data filter methods, the experimental results on 15 public PROMISE datasets show that DFAC increases PD value, reduces PF value and achieves higher G-measure and AUC values.

show abstract

“…The third mainstream way is to apply unsupervised classifier that does not require any training data to perform CCDP (e.g., [20][21][22][23]), therefore the distribution gap between the training project data and the target project data is no longer an issue. For instance, Zhang et.al [22] proposed to apply a connectivity-based unsupervised classifier that is based on spectral clustering to perform CPDP.…”

Section: A Defect Predictionmentioning

confidence: 99%

“…For instance, Zhang et.al [22] proposed to apply a connectivity-based unsupervised classifier that is based on spectral clustering to perform CPDP.…”

Section: A Defect Predictionmentioning

confidence: 99%

Combing Data Filter and Data Sampling for Cross-Company Defect Prediction: An Empricial Study

Yu

¹

,

Wu

²

,

Zhang³

et al. 2017

International Conferences on Software Engineering and Knowledge Engineering

0

View full text Add to dashboard Cite

Abstract-Cross-company defect prediction (CCDP) is a practical way that trains a prediction model by exploiting one or multiple projects of a source company and then applies the model to target company. Unfortunately, larger irrelevant crosscompany (CC) data usually makes it difficult to build a prediction model with high performance. On the other hand, the CC data has the highly imbalanced nature between the defectiveprone and non-defective classes, which will degrade the performance of CCDP. To address such issues, this paper proposes an approach, in which data sampling is combined with data filter, to overcome these problems. Data sampling seeks a more balanced dataset through the addition or removal of instances, while data filter is a process of filtering out the irrelevant CC data so that the performance of CCDP models can be improved. We employ two data filtering methods called NN filter and DBSCAN filter combined with SMOTE (Synthetic Minority Oversampling Technique) and RUS (Random UnderSampling). Eight different approaches would be produced when combing these four techniques: 1-NN filter performed prior to RUS; 2-NN filter performed after RUS; 3-NN filter performed prior to SMOTE; 4-NN filter performed after SMOTE; 5-DBSCAN filter performed prior to RUS; 6-DBSCAN filter performed after RUS; 7-DBSCAN filter performed prior to SMOTE; 8-DBSCAN filter performed after SMOTE. The empirical study was carried out on 15 publicly available project datasets. The experimental results demonstrate that NN filter performed prior to RUS (Approach 1) performs better than the other seven approaches.

show abstract

Cross-project defect prediction using a connectivity-based unsupervised classifier

Cited by 178 publications

References 52 publications

Heterogeneous fault prediction with cost‐sensitive domain adaptation

Heterogeneous fault prediction with cost‐sensitive domain adaptation

A Data Filtering Method Based on Agglomerative Clustering

Combing Data Filter and Data Sampling for Cross-Company Defect Prediction: An Empricial Study

Contact Info

Product

Resources

About