Test case generation techniques based on isolation forest algorithms

Abstract

Test cases are the key to software testing, and it is particularly important to generate and select test cases that detect faults more quickly. This paper first analyses the characteristics and applicability of typical ART, then combines the isolation forest algorithm with ART, proposes a test case generation technique for IForest-ART, and compares the fault detection effects of typical distance-based ART and IForest-ART in low to high dimensions. The experimental results show that the test validity of IForest-ART in high-dimensional input domain space is significantly higher than that of FSCS-ART, which effectively alleviates the problem of decreasing test validity of typical ART in high-dimensional input domain space and provides a new idea for ART.