A cross -version depth deficiency prediction method that can alleviate overlap problems

Abstract

The invention discloses a cross-version depth defect prediction method that can alleviate the problem of class overlap, including the following: 1. An overall framework for deep semantic learning in cross-version software defect prediction; 2. A semantic feature learning model based on a convolutional neural network 3. Hybrid nearest neighbor cleaning strategy for deep semantic learning. The present invention adopts a hybrid nearest neighbor cleaning strategy to alleviate the class overlap problem existing in the semantic features learned by deep learning. Specifically, for the abstract syntax tree corresponding to the source code, a convolutional neural network is used to learn deep semantic features, and then a hybrid nearest neighbor cleaning strategy is used to resample and clean the labeled dataset. The hybrid nearest neighbor cleaning strategy can deal with the problem of class imbalance and class overlap. The results of statistical analysis of the data show that this strategy can improve the performance of software defect prediction based on deep semantic learning.

Description

technical field [0001] The invention specifically relates to a cross-version depth defect prediction method which can alleviate the problem of class overlap. Background technique [0002] Software defect prediction is used to identify software defects in the software development process. The historical data generated in the software development process constitutes the training data of the software defect prediction classifier, and these data can be labeled from multiple granularities such as files and classes. Historical data-oriented metrics from the software process perspective are manually designed to build classification models, including lines of code-based metrics, Halstead scientific metrics, and McCabe loop complexities Degree (cyclomatic complexity) and so on. The traditional intra-project defect prediction model focuses on static metric elements and builds a classification model based on metric elements. Potential defective modules should have the same statistica...

AI Technical Summary

Problems solved by technology

However, these strategies are all based on traditional static metrics, and there are few studies on class overlap for semantic learning-based software defect prediction.

Images