The invention discloses a new class mining method and device. The method comprises the steps that when classification data is classified by using classification trees, the data sizes classified to nodes corresponding to target labels are recorded; the data sizes under the nodes corresponding to the target labels are monitored, and when the data sizes under the nodes corresponding to the target labels reach preset size values, the data under the nodes corresponding to the target labels with the data sizes reaching the preset size values is scanned and extracted and clustering calculation is performed to generate new subclass labels. By adopting the technical scheme, new class mining, namely further subdividing, is conducted on the data under the labels with too large data sizes, accordingly a classification system can be automatically updated, more accurate data classification can be provided, and the classification efficiency can be improved.