Software Fault-Prone Module Classification Using Learning Automata based Deep Neural Network Model

The software systems usually include some defects that will show adverse effects on the use of the application. Software fault-prone module will help software professionals to release the software within given time and cost. Software industries have to spend most of the budget in the testing phase o...

Full description

Saved in:

Bibliographic Details
Published in:	2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT) pp. 1 - 6
Main Authors:	Ahmed, Mohammad Mushtaq, Kiran, Bura Santhi, Sai, Pododdi Harshavardhan, Bisi, Manjubala
Format:	Conference Proceeding
Language:	English
Published:	IEEE 06-07-2021
Subjects:	Costs Deep learning Deep neural network Learning automata Predictive models Sampling methods Software algorithms Software defect prediction Software metrics Software systems
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	The software systems usually include some defects that will show adverse effects on the use of the application. Software fault-prone module will help software professionals to release the software within given time and cost. Software industries have to spend most of the budget in the testing phase only. If the fault-prone software modules are identified before testing phase, the limited testing resources will be allocated to the fault-prone modules to reduce the cost. In this paper, we have proposed a novel approach that leverages deep learning techniques to classify the software modules into faulty and non-faulty modules using learning automata. The deep learning model is trained using learning automata. We have validated our proposed approach using publicly available data set taken from public repository. We have also applied over sampling techniques on four data sets to balance the data set in order to obtain better performance measures. We have compared the results of proposed approaches with some existing models and we found our proposed approach is able to provide better prediction performance.
DOI:	10.1109/ICCCNT51525.2021.9580173