Comparison of Random and Partition Testing Considering the Test Profile

Abstract

Abstract There has been much interest in the analysis of the effectiveness of partition and random testing. Investigators studied and compared these two strategies from different points of view and came to different conclusions. Some used the probability of detecting at least one failure to quantify the fault-detecting ability of a testing strategy. Some used the expected number of failures detected to assess them. But they did not consider the test profile. Differing from this early work, this paper compares random testing with partition testing by looking at three parameters: the confidence index, the upper confidence bound, and the total number of test cases, considering the test profile. Our results show that random testing generally performs better than partition testing and equally as well as partition testing when the operational profile coincides with the test profile. We also derive the formulas to calculate the three indexes for random and partitional testings.