Background: Multi-source cross-project defect prediction (MSCPDP) attempts to transfer defect knowledge learned from multiple source projects to the target project. MSCPDP has drawn increasing attention from academic and industry communities owing to its advantages compared with single-source cross-project defect prediction (SSCPDP). However, two main problems, which are how to effectively extract the transferable knowledge from each source dataset and how to measure the amount of knowledge transferred from each source dataset to the target dataset, seriously restrict the performance of existing MSCPDP models. Objective: In this paper, we propose a novel multi-source transfer weighted ensemble learning (MASTER) method for MSCPDP. Method: MASTER measures the weight of each source dataset based on feature importance and distribution difference and then extracts the transferable knowledge based on the proposed feature-weighted transfer learning algorithm. Experiments are performed on 30 software projects. We compare MASTER with the latest state-of-the-art MSCPDP methods with statistical test in terms of famous effort-unaware measures (i.e., PD, PF, AUC, and MCC) and two widely used effort-aware measures (P_opt 20% and IFA). Result: The experiment results show that: 1) MASTER can substantially improve the prediction performance compared with the baselines, e.g., an improvement of at least 49.1% in MCC, 48.1% in IFA; 2) MASTER significantly outperforms each baseline on most datasets in terms of AUC, MCC, P_opt 20% and IFA; 3) MSCPDP model significantly performs better than the mean case of SSCPDP model on most datasets and even outperforms the best case of SSCPDP on some datasets. Conclusion: It can be concluded that 1) it is very necessary to conduct MSCPDP, and 2) the proposed MASTER is a more promising alternative for MSCPDP.

Imbalanced data are a major factor for degrading the performance of software defect models. Software defect dataset is imbalanced in nature, i.e., the number of non-defect-prone modules is far more than that of defect-prone ones, which results in the bias of classifiers on the majority class samples. In this paper, we propose a novel credibility-based imbalance boosting (CIB) method in order to address the class-imbalance problem in software defect proneness prediction. The method measures the credibility of synthetic samples based on their distribution by introducing a credit factor to every synthetic sample, and proposes a weight updating scheme to make the base classifiers focus on synthetic samples with high credibility and real samples. Experiments are performed on 11 NASA datasets and nine PROMISE datasets by comparing CIB with MAHAKIL, AdaC2, AdaBoost, SMOTE, RUS, No sampling method in terms of four performance measures, i.e., area under the curve (AUC), F₁, AGF, and Matthews correlation coefficient (MCC). Wilcoxon sign-ranked test and Cliff’s $\delta$ are separately used to perform statistical test and calculate effect size. The experimental results show that CIB is a more promising alternative for addressing the class-imbalance problem in software defect-prone prediction as compared with previous methods.

Line-level software defect prediction (LLDP) is crucial for locating and modifying defective code and has drawn increasing attention from both academic and industrial communities. Recently, a series of deep learning-based LLDP models have been proposed. However, there is no systematic comparison of these LLDP models. In this paper, we aim to compare the performance of these models, evaluate their consistency in different scenarios, and investigate whether fusing handcrafted features can improve the prediction performance. To this end, we conducted a comprehensive empirical study on eight recent state-of-the-art models using 32 benchmark datasets from nine software projects’ releases. We evaluated them in both cross-version and cross-project scenarios with four widely used performance metrics, including AUC, Recall@Top20%LOC, Effort@Top20%Recall, and IFA. Moreover, we investigated the impact of handcrafted features on LLDP models based on two different fusion strategies. The results show that 1) the difference among these models is statistically significant; 2) no model is always superior across all performance metrics but Bugsplorer generally performs best except IFA; 3) model performance rankings are inconsistent between cross-version and cross-project scenarios, revealing that a model’s effectiveness is highly scenario-dependent; 4) furthermore, fusing handcrafted features significantly improves model prediction performance, and the fusion strategy also matters. In conclusion, the selection of an LLDP model should be guided by specific scenarios and performance metrics, and the hybrid model that combines deep learning representations with handcrafted features is a promising alternative for LLDP.