Nuclear forensics is one of the activities to classify the sources of unidentified nuclear materials. Korea Institute of Nuclear Nonproliferation and Control (KINAC) has been conducting researches on this field continuously and the data-driven approaches have been studied as a forensic method for fresh and spent nuclear fuels in recent years. Under this circumstance, this article summarizes the methodologies implemented at KINAC and Kyung Hee University. It includes (1) study on identifiable signatures, (2) development of nuclear forensic library, and (3) research on classification algorithm for the identifiable signatures. There are various characteristics such as physical, chemical, and radiological properties, which are considered as the identifiable signatures of nuclear fuels. First, the fresh or spent fuel can be classified according to their radiation level. The impurities can be considered as the main signature for fresh fuels in case that they have generally a similar level of enrichment. The overall characteristics of multi-dimensional impurity dataset are extracted through Principal Component Analysis (PCA) and finally one-class Support Vector Machine (SVM) is used to classify the reactor types in which the fuel was used. In spent fuel forensics, the radionuclide mass through radiation analysis is focused on as a signature. The application of the classification of the reactor type using SVM and the regression analysis to predict the operational history, such as enrichment, burn up, and cooling time are investigated. The potential of data-drive methodologies for nuclear forensic is discussed.