Adaptive Random Test Case Generation Based On The Density Of Grid Region

Software testing strives to generate and select representative test cases to detect failures in software programs at lower cost and faster efficiency.In the existing test case generation methods,random testing(RT)selects test cases based on the characteristics of random distribution.This method has been widely used in practical testing because of its simplicity,but it cannot effectively use the information of spatial location,so the effect of failure-detection is not remarkable.Researchers have proposed adaptive random testing in order to improve the failure-detection ability of random testing,and this method uses location information of test cases that has been executed but found no faults,so that test cases can be more "uniformly" distributed in the domain and bring significant effect of failure-detection.As a typical method for adaptive random testing,Fixed-Sized-Candidate-Set ART has been well studied because of its good effect since it was proposed.However,as the dimension of the input domain increases,the test cases it generates are largely close to the edge of the input domain,which leads to poor failure-detection effect,and the huge calculation makes its efficiency drop extremely.Aiming at the problems of poor failure-detection effect and low efficiency in FSCS-ART algorithm,this thesis proposes a new adaptive random test case generation algorithm named ART-DGR(Adaptive Random Testing by Density of Grid Region).It uses dynamic griding to partition the input domain,so that the generated test cases appear more uniform.Considering the closeness of regions and the distance of test cases,a region candidate strategy and a distance candidate strategy are used to double-screen the candidate targets of the algorithm.And the region density represents the degree of close connection between the sub-region and the entire input domain.This method improves the failure-detection effect of FSCS-ART algorithm.A forgetting strategy based on Manhattan distance is proposed which greatly improves the computational efficiency,thereby solving the efficiency of FSCS-ART algorithm.Finally,detailed simulation experiments and empirical experiments are carried out to verify the effectiveness of ART-DGR algorithm.The experimental results show that the algorithm designed in this thesis meets the requirements of "uniform distribution" and has an ideal spending of calculation.Compared with FSCS-ART algorithm,it has a better failuredetection effect.The improved ART-DGR algorithm also shows faster efficiency in the high-dimensional input domain.Test case automatic generation helps testers to detect program defects efficiently,which can greatly reduce human resources cost and time consumption.Through deeply investigating the existing ART algorithms,we propose a novel ART algorithm,namely ART-DGR,which applies the concept of region density to spread test cases in input domains.At the same time,we also design a solution for the high-dimensional cases.It is easy to see that,the proposed ART algorithm has important reference significance to the software testing academic field and the industrial practical test field.