Cross-project software defect prediction method based on shared hidden layer auto-encoder

Abstract

The invention discloses a cross-project software defect prediction method based on a shared hidden layer auto-encoder, and the method comprises the steps: firstly preprocessing a data set, and dividing the data set into a training set and a test set; secondly, performing feature extraction by adopting an auto-encoder with a sharing mechanism, and respectively extracting depth features of the training set and the test set; finally, introducing a focus loss function to train and a classifier. The problem of feature distribution difference in cross-project software defect prediction is solved, and a focus loss-based shared hidden layer auto-encoder technology is proposed for the first time. Different data distributions become more similar, different weights are allocated to different types of samples by using a focus loss learning technology to solve class imbalance, and meanwhile, different weights are given to samples easy to classify and samples difficult to classify, so that a classifier can better learn the samples difficult to classify.

Description

technical field [0001] The invention belongs to the field of software engineering, and in particular relates to a cross-project software defect prediction method based on a shared hidden layer autoencoder. Background technique [0002] Software defect prediction is a research hotspot in the field of software engineering. Its main goal is to discover defects in software in advance in the early process of software development and improve the quality of software products. Most previous research has focused on the problem of intra-project defect prediction, mainly using a part of the historical data of the same project to train a prediction model, and then using the remaining data of the same project to test the ability of the model to predict defects. But for a newly started project, there is not enough historical data to train the model, and the performance of defect prediction within the project will perform poorly. Therefore, when there is not enough historical defect data...

AI Technical Summary

Problems solved by technology

However, its prediction performance is still very poor. The main reason is that there is a difference in data distribution between the source project and the target project. If the difference in their data distribution is smaller, the cross-project defect prediction effect is better.

In addition, due to the class imbalance in the data set itself, that is, the number of non-defective categories is much greater than the number of defective categories, this class imbalance problem will reduce the prediction performance of the model, making it easier for the model to identify samples of non-defective categories, Shows poor predictive power for flawed samples

Images