Defect Prediction Technology of Aerospace Software Based on Deep Neural Network and Process Measurement

Tianwen Yao,Bo Zhang,Ben Zhang,Zhiqiang Han,Jun Peng,Zhi Zhang,Zhaobing Yang,Antonio M Gonçalves De Lima

doi:10.1155/2022/1276830

Abstract

In order to ensure high reliability, the efficiency of traditional aerospace software testing is often low. With the rapid development of machine learning, its powerful data feature extraction ability has great potential in improving the efficiency of aerospace software testing. Therefore, this paper proposed a software defect prediction method based on deep neural network and process measurement. Based on the NASA data set and combined with the software process data, the software defect measurement set is constructed. 35 measurement elements are used as the original input, and multiple single-layer automatic coding networks are superimposed to form the deep neural network model of software defect. The model is finally trained by the layer-by-layer greedy training method to realize software defect prediction. Experimental verification shows that the prediction method has a good prediction effect on aerospace software defects, and the accuracy rate reached 90%, which can greatly improve the efficiency and effect of aerospace software testing.

Highlights

With the continuous development of aerospace technology, the scale and complexity of aerospace software are getting higher and higher, and the development cycle of aerospace models is gradually shortening. is poses new challenges to the development process of software in the aerospace field
In order to improve the efficiency of aerospace software defect model, this paper analyzed the characteristics of aerospace software and used machine learning methods to study the defect prediction of aerospace software and proposed a software defect prediction method based on deep neural networks and process metrics
A single-layer network was constructed through data dimension reduction, but the depth is too shallow, and it will lead to high redundancy of the data feature; a good prediction effect that is desired cannot be achieved. erefore, this paper proposes to use the self-encoding network model to form a deep neural network model to further reduce the dimension and learn the features, which can greatly improve the classification ability and improve the prediction accuracy

Summary

Introduction

With the continuous development of aerospace technology, the scale and complexity of aerospace software are getting higher and higher, and the development cycle of aerospace models is gradually shortening. is poses new challenges to the development process of software in the aerospace field. Since the 1970s, researchers have begun to build metric set based on the software history warehouse and used statistics and machine learning methods to study software defect prediction. When they trained the model, they were based on the training data of the previous software version, and on selecting and marking a small number of examples from the current software version It can be seen from current researches that when constructing a software defect prediction model, it is necessary to select an appropriate machine learning method according to the characteristics of the predicted object. In order to improve the efficiency of aerospace software defect model, this paper analyzed the characteristics of aerospace software and used machine learning methods to study the defect prediction of aerospace software and proposed a software defect prediction method based on deep neural networks and process metrics. E main research in the content includes designing a set of software defect metrics for the aerospace field and constructing a deep neural network model for aerospace software defect prediction

Metric Set for Aerospace Software

Construction Strategy of Aerospace Software Defect Prediction Model

Experimental Results and Analyses

A1-2 A1-3 A2-3 A1-2-3

Conclusion