This work focused on evaluating some machine learning models for their application in e-Health IoT platforms used to detect epileptic seizures. The evaluation is based on two groups of metrics: statistical validation and computational complexity. These metrics determine relevant factors for the models' selection based on the intrinsic efficiency of ML models to detect seizures and their IoT appropriateness to reduce computation and, therefore, energy use. The evaluation scenario defines an EFC architecture with Edge, Fog, and Cloud layers, where the models are deployed initially in the cloud layer for training and validation, to later be deployed in the fog and edge layers for use. Results highlight that GBC and XGBC models present better performance when executed from the cloud; LR, NB and SNN models can be trained from fog nodes, and finally, SLR and MLP can be deployed and used from edge nodes. MLP especially presents a good balance between a low computational cost and a high accuracy in seizure detection.