Test Case Recommendation for Blockchain Systems Based on Graph Convolutional Networks

In the development and operation of blockchain systems, the evaluation of these systems typically necessitates the formulation and execution of test cases. Current testing practices for blockchain systems often involve the manual specification of relevant test scenarios, with testing being conducted through inquiries or the execution of tasks within the system. However, due to the relatively narrow scope of real-world applications for blockchain systems, there exists a deficiency in appropriate test case sets for the evaluation of such systems. Consequently, this paper presents a Blockchain Test Case Recommendation System based on LightGCN, a simplified Graph Convolutional Neural Network. LightGCN dispenses with traditional graph convolutional network components such as non-linear activation and feature transformations, thereby enhancing the efficiency of model training. Experiments illustrate that this system effectively supplements and enhances the blockchain system’s test case set by aggregating information from domain nodes.