Feature selection is regarded as an important task in data mining. The applications of machine learning eliminate irrelevantly, redundant features so that the learning performance is improved. A novel feature selection method for unsupervised text clustering, that is, binary multi-verse optimizer algorithm (BMVO) is proposed in this paper. A new application of the MVO algorithm is introduced via this method, which selects important text features. Then, these important features are tested using a k-means clustering algorithm to enhance performance and lessen the cost of the proposed algorithm computational time. The BMVO performance is examined on 6 datasets that are published including Classic4, Wap, tr41, tr12, 20Newsgroups, and CSTR. Based on the measures of the evaluation, the obtained results showed that the BMVO performance has outperformed the rest of the comparative algorithms. © 2019 IEEE.