Software defect prediction method based on neural network

Abstract

The invention discloses a software defect prediction method based on a neural network; the method comprises the steps: software feature preprocessing: dividing the feature dimension of tested software into N equal parts, and enabling the data of each equal part to obey the normal distribution N (mu, sigma<2>); sampling: firstly, calculating a median of a value of each tested software sample under each software feature, and taking the median of each software feature as a threshold value of the software feature, for each sample, comparing the value of each software feature with a corresponding threshold value, calculating the number of the features, greater than the threshold value, of the tested software sample, and sorting the samples from more to less; and sampling the samples according to the sequence, so that defect data occupies a certain proportion; setting a neural network model: enabling a target value to be equal to an input value as much as possible by using a back propagation algorithm, adjusting neural network parameters through continuous training, and obtaining a training weight of each layer of neural network so as to achieve the purpose of feature dimension reduction. According to the invention, the software defect discovery efficiency can be remarkably improved.

Description

technical field [0001] The invention relates to the technical field of software testing, in particular to a neural network-based software defect prediction method. Background technique [0002] Software defects refer to problems such as errors in the development or maintenance process of software products, or the failure or violation of certain functions that need to be realized. With the increasing scale and complexity of software, defect data will inevitably be generated in the software development process, and defects may appear in any stage of the software development process. With the development of software technology so far, it is impossible to find and eliminate all defects by any means of inspection and verification. In the development cycle of a software project, the later a defect is discovered, the higher the cost of repairing the defect will be. [0003] Statistics show that in the software code review stage, it takes two minutes to find and repair a software d...

AI Technical Summary

Problems solved by technology

[0003] Statistics show that in the software code review stage, it takes two minutes to find and repair a software defect evaluation, ten to twenty minutes in the unit test stage, an hour in the integration test stage on average, and an hour in the system test stage. Forty hours, and the cost of fixing bugs grows almost exponentially over time

Images