SWFC-ART: A cost-effective approach for Fixed-Size-Candidate-Set Adaptive Random Testing through small world graphs

Test case generation techniques based on isolation forest algorithms

“…Random testing (RT) is often used as one of the common methods for test case generation due to its simple, efficient and easy-to-implement algorithm, but at the same time, random testing also tends to lead to problems such as redundancy and low coverage of test cases [1].In recent years, an improved adaptive random testing (ART) method has been proposed to improve the test coverage of random tests, as shown in Figure 1,which has received a high level of attention and has spawned a variety of ARTs with different strategies, mainly divided into four major categories: distance-based ART [2], restriction-based ART [3], division-based ART [4] [5], and other ART(grid-based ART) [6]. The above methods have significantly improved fault detection compared to RT, but they all face many problems, such as high time costs, significant boundary effects and unsuitable high-dimensional input domains.Rubing Huang et al proposed an adaptive small-world graph-based stochastic test method (SWFC-ART) to improve the computational efficiency of FSCS-ART by reducing the computational overhead of FSCS-ART from quadratic to log-linear order while maintaining the fault detection efficiency of FSCS-ART with consistency in the high-dimensional input domain [7].Mengting Quan et al proposed the FSCS-ART algorithm based on a central compensation strategy in order to address the boundary effect and computational efficiency of FSCS-ART [8].In response to two types of problems in the FSCS-ART method, namely poor fault detection and low operational efficiency, Zhilei Chen et al proposed ART-DGR, an adaptive random test case generation algorithm based on grid area density [9].Chen et al proposed a distance-aware forgetting strategy for fixed candidate set size ART (DF-FSCS), which takes into account the spatial distribution of test cases and ignores test cases outside the "line of sight" of a given candidate in order to reduce the distance computation cost. It also uses dynamic adjustment of partitioning and second round of forgetting to ensure the linear complexity of the DF-FSCS algorithm [10].Chen et al also proposed a new test case generation method based on iterative partitioning, which reduces the overhead of test case generation by dividing the input field into equal cells [11].C.…”

“…(6) Use isolated forest to find test cases that are easily isolated. (7) Keep the candidate test cases that are easily isolated in the selected test case set, and remove the rest from the selected test case set. (8) determine whether the selected test case set can detect faults, if it can, go to (9), otherwise go to (5).…”