Autor: |
Juneja, Sapna, Nauman, Ali, Uppal, Mudita, Gupta, Deepali, Alroobaea, Roobaea, Muminov, Bahodir, Tao, Yuning |
Předmět: |
|
Zdroj: |
Journal of Supercomputing; May2024, Vol. 80 Issue 7, p10122-10147, 26p |
Abstrakt: |
When it comes to software development, precise planning, proper documentation and proper process control, some errors are inevitable in the software environment. These software flaws can lead to quality deterioration, which can be the main reason behind system failure. As the whole world especially developing countries is dependent upon software systems, it is very important to focus on its reliability aspect. Nowadays sophisticated systems require concerted efforts for managing and reducing the shortcomings in software engineering. But, these efforts require more cost, more money and more time. Software error prediction is the most helpful step in the testing stage of the software development life cycle. It identifies components or parts of the code where an error may occur and requires broad testing, so the test resources can be efficiently used. Software error assessment reduces efforts of testing the software by helping the software testers locate the actual problem and classify different classes of errors in the system. Error estimators are majorly used in various organizations to evaluate the software to save time, improve the quality of software and testing and optimize resources to meet timelines. Machine learning provides support in fault projection by collecting the training data from various edge devices and thus helps in escalating the reliability of the software available on Kaggle. The multilayer perceptron shows better results in precision, recall, F1 score and accuracy as compared to decision tree and Gaussian Naive Bayes as it achieves an accuracy of 96.8%. [ABSTRACT FROM AUTHOR] |
Databáze: |
Complementary Index |
Externí odkaz: |
|