Test case generation optimization using combination prediction model and prioritization identification for complex system retesting

Abstract

Nowadays, the retesting process has become crucial in assessing the functionality and correctness of a system in order to ensure high reliability. Although many techniques and approaches have been introduced by researchers, some issues still need addressing to ensure test case adequacy. To determine test case adequacy, it is crucial to first determine the test set size in terms of number of test cases to prevent the system from failing to execute. It is also crucial to identify the requirement specification factor that would solve the problem of insufficiency and scenario redundancy. To overcome this drawback, this study proposed an approach for test case generation in the retesting process by combining two models, which would reveal more severe faults and improve software quality. The first model was enhanced through determining the test case set size by constructing a predictive model based on failure rate using seed fault validation. This model was then extended to requirement prioritisation. Next, it was used to schedule the test cases that focus on Prioritisation Factor Value of requirement specifications. The Test Point Analysis was used to evaluate test effort by measuring level of estimation complexity and by considering the relationship among test cases, fault response time, and fault resolution time. This approach was then evaluated using complex system that called as Plantation Management System as a project case study. Data of Payroll and Labour Management module that applied in 138 estates been collected for this study. As a result, the test case generation approach was able to measure test effort with High accuracy based on two combination model and it achieved a complexity level with 90% confidence bounds of Relative Error. This result proves that this approach can forecast test effort rank based on complexity level of requirement, which can be extracted from early on in the testing phase.