Improving the methods used for the detection and estimation of fissile material mass content is essential to nuclear security and safeguards to ensure special nuclear material (SNM) accountability, control, safety and security. With the continued expansion of the nuclear industry and the need for safe management of spent fuel, improving existing nondestructive assay (NDA) techniques is imperative. In the present research, we are particularly interested in Pu-239 and U-235 content. Following thermal neutron-induced fission, temporal gamma-ray spectroscopy and machine learning methods were used to classify pure samples of Pu-239 and U-235. Further, regions of interest identified during data pre-processing were utilized for computing the relative mass content of the fissile materials. The temporal gamma-ray spectroscopy method takes advantage of the time-dependent decay of fission products. Without prior knowledge of peak locations or their associated energies, temporal patterns characteristic of radioactive decay were identified within the complex fission product gamma-ray spectra below 3 MeV. Following feature generation and feature selection, Welch’s t-test was employed to tease out regions of interest used as "fingerprints" to create profiles of Pu-239 and U-235 to be used as input into four machine learning architectures: decision tree, random forest, neural network and Bayesian network. Classification accuracy ranged from 98-100% for all four classifiers. Initial results for determining relative mass content are promising as several regions of interest showed mass estimates within two standard deviations uncertainty for at least one of the fissile materials.