By classifying crops using machine learning approaches, it is possible to determine if the spectral signatures of several variations of the same species differ from one another. This allows for the correlation of the spectral signatures with key properties of the finished product. The final cannabinoid content of a certain species is a crucial quality attribute that might raise the crop’s value in the case of Cannabis growing. In contrast to conventional cutting and laboratory analysis approaches, the classification of Cannabis varietals from spectral signatures is proposed as a nondestructive process. The findings demonstrate that a random forest classification algorithm optimized on hyparameters can classify four types of Cannabis grown in Colombia with a multiclass accuracy of 95.6% using the spectral signature. These findings will make it possible to determine whether the spectral signature is related to the cannabinoid content of the various kinds, which is crucial for medical purposes.