Software Fault Prediction Analysis under BPSO Dimension Reduction Conditions
Download as PDF
DOI: 10.25236/csam.2019.078
Corresponding Author
Liu Hongqing
Abstract
The identification of module fault tendency is very important to reduce cost and improve the effectiveness of software development process. The software fault tendency module deep neural networks (DNN) prediction based on the bound particle swarm optimization (BPSO) dimensionality reduction is put forward. Firstly, the calculation framework of the BPSO dimensionality reduction-based software fault tendency module DNN prediction algorithm and the measure indexes of 21 software faults used are given, and the normalized preprocessing method of its index values is also given; secondly, the dimensionality reduction is performed on the software fault dataset by bound particle swarm optimization, and the participle position is represented by a binary (0 or 1) character string to simplify the data processing. Then the prediction of software fault tendency module is realized by the deep neural network algorithm; finally, the performance advantages of the algorithm are verified by simulation experiments on the four standard test sets of PC1, JM1, KC1 and KC3.
Keywords
Bound Particle Swarm Optimization, Software Fault, Deep Neural Network, Dimensionality Reduction, Bound State