【Background】Several major pests in agriculture and forestry belong to the lepidopteran family of Noctuidae, which are often difficult to distinguish morphologically. As a fast and accurate approach to the identification of species, DNA barcoding is used in identification since 2003. This, accompanied with a machine learning method, the support vector machine (SVM) has achieved good results in data classification, high dimensional pattern recognition and other fields. The combination of these two methods holds promise to ease the correct identification of noctuids, which would improve pest control efforts. 【Method】DNA barcodes of the mitochondrial cytochrome oxidase c subunit Ⅰ (COⅠ) gene, divided into three sets of databases, were tested by the neighborjoining (NJ) method and the SVM. We used 101 specimens of 58 species of Noctuidae which were identified by experts. These specimens were collected mainly from the Miaofeng Mountain, north of Miaofengshan Town, Beijing, China. 【Result】The NJ method proved to be superior to the SVM. However, some important species, such as Helicoverpa armigera and H.assulta, could not be distinguished in the NJ-tree. 【Conclusion and significance】As a new technology of species identification, DNA barcoding can provide assistance in problems in taxonomy; combination with the NJ method gave better support than the SVM.