Automatic directory classification of test cases based on Machine Learning Algorithms at an Android Smartphone Vendor

Abstract

Software test cases is an important study issue that has piqued the interest of many academics who are attempting to create or suggest a heuristic strategy that might lessen the laborious manual effort that software engineers expend while classifying test cases. The goal is to ensure that all features and apps have been tested and verified. In order to achieve that, there must be a good framework that can suggest or match the feature labels with their test cases in a chronological way. Failing to do so will result in inaccurately labeled test cases. Therefore, the key objective of this paper is to propose a method that can do an automatic directory classification of test cases based on their test case description by applying the K-nearest neighbor classifier. Bag-of-word (Bow) and Term Frequency-Inverse Document Frequency were used as a vector representation and fitted the KNN classifier. The experimental result shows that using KNN-BOW has a good score compared to KNN-TF-IDF as it outperformed and achieved 77% accuracy in comparison with the 71% that KNN-TF-IDF achieved. Because of that, KNN-BOW is a good option for the directory classification based on test case descriptions. The proposed method has a contribution to the domain and makes sure that using machine learning algorithms can make easy directory classification of test case descriptions.

Keywords

• test cases
• k nearest neighbor
• bag-of-words
• machine learning algorithms