Abstract: In order to improve the effectiveness of fault localization,various test suite reduction and selection methods have been proposed in recent years.However,excessive reduction or improper selection on test cases may result in the loss of some testing information and affect the fault localization.In this paper,a new test suite selection approach is proposed to improve spectrum-based fault localization (SFL).The proposed approach can eliminate the deviation in the accuracy of fault localization caused by coincidental passed test cases.By selecting similar test cases for each failed test case from the past test set,testing information is retained as more as possible.Besides,a ranking list is constructed by using an SFL technique with the new spectra.Finally,it is shown via experimental results on Siemens programs that the proposed approach can improve fault-localization effectiveness significantly.