Software testing has an essential role in a software development life cycle. Once a software development is implemented; software developers dedicate considerable effort in the testing process to improve the software reliability, quality and performance. This process includes the test cases planning, generating test inputs for exploring behaviors of the program, etc. The time consumed on testing is primarily concerned with testing great numbers of unreliable test cases. The primary goal here is to decrease these numbers and propose more reliable test cases, and this is feasible by using certain techniques selection to determine a subset of existing test cases. The essential purpose of test case reduction is to reduce the testcases which are redundant and in the same time maintaining the efficiency and effectiveness of the testcases. Regression testing is a major component of the software testing life cycle. It is conducted to make sure that the previously working functionality is not affected by the changes in the software environment occurred by the fixes or any enhancement changes. This thesis aims to offering a new approach in testcase reduction to reduce the number of testcases by eliminate the redundant testcases, for that purpose,  four datasets have been  built, each of which consists of several testcases and it’s attributes, then we applied several data mining techniques on these datasets from various categories such as canopy, k-mean, farthest first, filtered cluster and make density cluster from clustering techniques. Out of classifier techniques, naïve base, K-nearest neighbor, random forest and decision tree have been applied, and from ensemble method we applied bagging, boosting and stacking. The results demonstrated that ensemble paradigm is capable of improving the testing process, which ultimately make improvement in the process of test case reduction.​​