Liquid biopsy refers to the sampling, screening, and detecting potential biomarkers in unique liquid samples for clinical use. Lung cancer is one of the most highly frequent cancer subtypes, which is hard to be early diagnosed and monitored by radiological and histopathological evaluation that are the most general and accurate methods. Circulating miRNA is a potential clinical examination index for tumor detection and monitoring tumorigenesis progression using liquid biopsy. However, recognizing and validating the unique clinical values of each candidate circulating miRNA is expensive and time consuming. In this study, we presented a novel computational approach for identifying significant circulating miRNAs that may be applied to early screening, diagnosis, and constant monitoring of lung cancer progression. This approach incorporated several machine learning algorithms and was applied on the expression profiles of circulating miRNAs on lung cancer patients and control samples. In brief, a powerful feature selection method, minimum redundancy maximum relevance, was adopted to evaluate the importance of all features, resulting in a feature list. Then, incremental feature selection incorporating random forest followed to extract key circulating miRNAs. At the same time, an efficient classifier with MCC 0.740 was built. Top five circulating miRNAs, including miR-92a, miR-140-5p, miR-331-3p, miR-223, miR-374a, were analyzed and confirmed that they participated in the pathogenesis of lung cancer, indicating their significant prognosis power in lung cancer.Copyright © 2020. Published by Elsevier B.V.