Empirical Studies of a Two-Stage Data Preprocessing Approach for Software Fault Prediction

Abstract

Software fault prediction is a valuable exercise in software quality assurance to best allocate limited testing resources. Classification is one of the effective methods for software fault prediction. The classification models are trained based on the datasets obtained by mining software historical repositories. However, the performance of the models depends on the quality of datasets. In this paper, we propose a novel two-stage data preprocessing approach which incorporates both feature selection and instance reduction. Specifically, in the feature selection stage, we first perform relevance analysis, and then propose a threshold-based clustering method, called novel threshold-based clustering algorithm, to conduct redundancy control. In the instance reduction stage, we apply random under-sampling to keep the balance between the faulty and non-faulty instances. In empirical studies, we chose datasets from real-world software projects, such as Eclipse and NASA. Then we compared our approach with some classical baseline methods, and further investigated the influencing factors in our approach. The final results demonstrate the effectiveness of our approach, and provide a guideline for achieving cost-effective data preprocessing when using our two-stage approach.

A Three-Stage Based Ensemble Learning for Improved Software Fault Prediction: An Empirical Comparative Study

“…To deal with these issues we have realized that the performance of ELA can be increased by keeping the quality of software defect datasets, which can be done by applying either FS, resolving class imbalance problem and/or filtering noise instances [12,13,18,25,26,41,42,44,47] from defect datasets. In this regard, Liu et al [45] made a comprehensive survey of FS algorithms.…”

“…However, there is no single best approach for all situations. Liu et al [47] proposed a twostage data preprocessing approach to improve the quality of software datasets used by classification models for SFP. In the FS stage, they proposed an algorithm which involves both relevance analysis and redundancy control.…”

“…Experimental results indicated that ensemble of very few rankers are effective and even better than ensemble of many or all rankers. As noticed, these studies didn't employ different ensemble techniques; perform DB (except [47]) and NF to resolve class imbalance issues and filter noise from the defect datasets.…”

“…Thereby, there is no significant difference between the number of samples in the majority and minority classes. However, when presented with complex imbalanced data sets, these algorithms fail to properly represent the distribution of the data and resultantly provide unfavorable accuracies across the classes of the data [2,5,11,[23][24][25][26]47]. Thus, in recent years, the imbalanced learning problem has drawn a significant amount of interest in Software Engineering research area, particularly in SFP.…”

“…Thus, to deal with these issues, an ensemble based combined framework has to be designed specifically. Therefore, in this study, our plan is to combine ELA with Feature Selection (FS) [9, 12-14, 20, 21, 45-47], Data Balancing (DB) [11,20,[23][24][25]47] and Noise Filtering (NF) [36][37][38][39][40][41][42]44] techniques. FS is carried out by removing less important and redundant features, so that only beneficial features are left for training the models.…”