Test Case Reuse Based on Software Testing Knowledge Graph and Collaborative Filtering Recommendation Algorithm

As an important role of software test, the reuse of test cases is essential in terms of finding software defects and locating the causes of them. However, the existing related approaches are insufficient to establish an internal relationship between test cases and defects and their abilities to find or diagnose errors are limited. In this paper, an ontology model based on the software testing process is applied to establish a software testing knowledge graph, which serves as the foundation to build an recommendation system. Specifically, the recommendation system takes the functions of software under test as the "user", and the defect-occurrence-chain which establishes the correlation between test cases and defects in the knowledge graph as the "item". Both of them provide the evidence to build collaborative filtering recommendation algorithm based on the user-item scoring matrix. It aims to assist testers in recommending reusable test cases to identify software errors effectively. Against this background, the BERT+Bi-LSTM-CRF model is selected to extract the latent test requirements of the software under test, and an overt variable factorization model is built so as to iteratively optimize the user-item scoring matrix. Further, an empirical study has been conducted, and it is found that the recommended test cases can significantly help testers find software defects faster in a more efficient way, and locate defects more accurately.