The cosmic-ray iron spectrum has been measured with the space-borne Cosmic Ray Energetics And Mass on the International Space Station (ISS-CREAM) experiment. The instrument was deployed on the ISS for 539 days, from August 22nd, 2017 to February 12th, 2019. A new machine-learning-based (ML) technique, using a convolutional neural network (CNN), was applied during the ISS-CREAM analysis. The iron spectrum reconstructed by the CNN is from 4.5 TeV to 73.9 TeV with a charge resolution of 0.7e (in charge units) and an energy resolution of ∼ 25%. A spectral index of 2.99 ± 0.20 was found by fitting the data with a power-law model. This estimate is consistent with the generally accepted value (∼ 2.60) within 2σ. However, our best-fit power-law normalization is ∼ 2 times lower than generally accepted. This difference is thought to result from the systematic uncertainty in the energy calibration methods of ISS-CREAM. In this thesis, the methods of ML and analysis of the iron flux will be discussed.