The complexity of current IC test is increasing, which leads to rising test cost. To address this issue, a reliable and effective test set optimization method is proposed in this article. The features in the original test set are filtered by clustering with K-means clustering. Then, a modified mRMR algorithm is used to introduce redundancy weighting factors among features by using a segmented formula to weigh the feature relevance and redundancy metrics. Meanwhile, the SVM cross-validation is inserted to reinforce the accuracy of the test pattern selection. The reduction in the number of dimensions of the original test set is achieved while ensuring that the fault coverage remains largely unchanged. Experiments on the ISCAS89 circuit show that the proposed method takes the test patterns of the original test set and streamlines them considerably. This method ensures test quality and also optimizes the test set. The test flow after redundancy elimination and sequencing reduces test time by 40. 43% , improves test efficiency and reduces test cost.