Utilizing machine learning algorithms to automatically categorize software test cases.

• Main Authors: bdirahma, Abdullahi Ahmed, Hashi, Abdirahman Osman, Mohd. Hashim, Siti Zaiton, Elmi, Mohamed Abdirahman
• Format: Article
• Language: English
• Published: Seventh Sense Research Group 2023
• Subjects: HD30.213 Management information systems. Decision support systems
• Online Access: http://eprints.utm.my/105616/1/SitiZaitonMohdHashim2023_UtilizingMachineLearningAlgorithstoAutomatically.pdf; http://eprints.utm.my/105616/; http://dx.doi.org/10.14445/22315381/IJETT-V71I9P203
• Institution: Universiti Teknologi Malaysia

The creation of an efficient strategy to help decrease the problem of manual effort expended by software developers when labelling software test cases has been the focus of many academic researchers. To ensure that all features and applications are fully tested, it is important to have a framework that can effectively match the feature labels and test cases in the correct sequence. Irrelevant labeling of test cases can result in inaccuracy, so avoiding it is a key objective of this paper. As a result, the primary goal of this work is to extend a previous method for doing automatic directory categorization of test cases based on their test case description using the K-nearest-neighbor classifier, Logistic regression, Decision tree and MLP. Bag-of-word (Bow) is used as a vector representation and fits all classifiers. The experimental results reveal that using KNN-BOW and MLP have a higher score than Logistic regression and Decision tree since it outperformed and obtained 77% accuracy vs. 71% for KNN-TF-IDF. Meanwhile, we extended using KNN-BOW and MLP-BOW have scored a good result compared to Logistic regression and Decision tree, as it outperformed and achieved 77% accuracy in comparison with the 67% and 65% that Logistic regression and Decision tree achieved, respectively. As a result, KNN-BOW and MLP-BOW are excellent choices for directory categorization based on test case descriptions. The suggested strategy contributes to the domain by ensuring that machine learning algorithms can easily directly classify test-case descriptions.