In 5G and beyond wireless communication systems, energy and spectral efficiency requirements should be satisfied while improving the error performance. Massive multiple input multiple output spatial modulation (MIMO-SM) systems are considered to be one of the candidate technologies for next-generation communication systems in terms of providing energy and spectral efficiency requirements. Error performance of massive MIMO-SM systems can be improved with Euclidean distance based antenna selection (EDAS), which strengthens this idea. In this paper, massive MIMO-SM systems are implemented for the first time in a real-time environment. In order to improve the error performance of the system, a machine learning based approach for transmitter antenna selection that has lower complexity than the optimal method. The designed system was on simulation and real-time environments. As a result of the study, in real-time systems nearest neighborhood (k-NN) algorithm's practicality has been demonstrated.