As mobile devices gain on popularity and see an increase in computing power as well as computational requirements, the limited energy capacity of the devices becomes more of a problem. A solution to this problem can be found in code offloading, where code computation is remotely executed in order to save energy and increase the processing speed. The most common offloading concept of cloud computing is ever more struggling to keep up with the demand of low latency and data privacy. This has started a shift in the computational paradigm where the computation centers are moved closer to the mobile devices and into the edge of the network. This thesis evaluates the efficiency of federated gradient boosted decision tree approaches, SimFL and FedTree, as classifiers for the offloading problem. The evaluation is based on real-world data about applications like decision tree classification, speech detection and photo filters. Multiple data distribution scenarios are tested to simulate a real-world scenario. The approaches are compared to a federated reinforcement learning approach as a performance comparison. The evaluation concludes, that currently FedTree is the most complete approach to the problem.