Pulse shape discrimination algorithms have been commonly implemented on embedded systems to discriminate neutron/gamma radiations detected by organic scintillators in several applications. These algorithms have a number of limitations, especially when used with plastic scintillators, which have low intrinsic discriminating ability. Machine learning (ML) models have recently been explored as a way to improve discriminating performance. Most of these methods are proposed for liquid and stilbene scintillators and do not address the integrated implementation. Reducing the sampling frequency of a discrimination system helps to minimize the size and cost of the embedded implementation. The purpose of this study is to explore whether the use of ML tools, compared to the Tail-to-Total integral ratio algorithm, can lead to a reduction in the minimum required sampling frequency in EJ276 plastic scintillator, as well as an enhancement in the classification performance. The results obtained highlight the superior performance of the ML model.