A Method for Predicting In-Project Software Aging Defects Based on Active Learning

Abstract

The invention discloses an Active Learning-based software aging prediction method in a project. By collecting static metrics of codes in the software, Active Learning is used to select samples and label them as a training set to predict the remaining samples without class labels. Active Learning is used for sample selection and manual labeling to form a training set. A combination of oversampling and undersampling is used to alleviate the class imbalance problem, and a machine learning classifier is used for prediction. The present invention considers that there are few samples in the software aging defect data set, and the collection is time-consuming and labor-intensive. The method of combining under-sampling and over-sampling is used to alleviate the problem of extreme class imbalance, which helps developers to find software aging-related defects during the development and testing stage. Removed to avoid losses caused by software aging issues. The feasibility of the invention has been verified on real software, and can be extended to other software to predict defects related to software aging.

Description

technical field [0001] The invention belongs to the technical field of software aging prediction, and in particular relates to an Active Learning-based in-project software aging defect prediction method. Background technique [0002] In long-running operating systems, software aging is a major cause of system performance degradation or software crashes. It is caused by software aging-related bugs (ARBs), such as memory leaks, unreleased file locks, storage problems, etc. And it has been found to exist in various systems, such as Android, Linux, Windows, etc. The complexity and time characteristics of software aging make its detection very difficult. Therefore, predicting and removing software aging-related defects in the development and testing stage (code level) is one of the important ways to reduce the loss caused by software aging. [0003] In recent years, aging defect prediction has gradually attracted the attention of scholars in the field of reliability. Some sch...

AI Technical Summary

Problems solved by technology

Although the amount of data in this method is sufficient, the differences between different projects are still relatively serious. Therefore, there are still certain differences between the prediction performance across projects and the prediction performance within projects.

Moreover, in previous studies, when dealing with extremely serious class imbalances, a single method of oversampling or undersampling was used, which can easily lead to overfitting and is not robust enough for different machine learning classifiers, that is, the prediction effect is quite different.

Images