The IceCube Neutrino Observatory is equipped with the unique possibility to measure cosmic ray induced air showers simultaneously by their particle footprint on the surface with the IceTop detector and by the high-energy muonic shower component at a depth of more than 1.5 km. Since 2019 additionally, two Imaging Air Cherenkov Telescopes, called IceAct, measure the electromagnetic component of air showers in the atmosphere above the IceCube detector. This opens the possibility to measure air shower parameters in three independent detectors and allows to improve mass composition studies with the IceCube data. One IceAct camera consists of 61 SiPM pixels in a hexagonal grid. Each pixel has a field of view of 1.5 degree resulting in an approximately 12-degree field of view per camera. A single telescope tube has a diameter of 50 cm, is built robust enough to withstand the harsh Antarctic conditions, and is able to detect cosmic ray particles with energies above approximately 10 TeV. A Graph Neural Network (GNN) is trained to determine the air shower properties from IceAct data. The composition analysis is then performed using Random Forest Regression (RF). Since all three detectors have a different energy threshold, we train several RFs with different inputs, combining the different detectors and taking advantage of the lower energy threshold of the IceAct telescopes. This will result in composition measurements for different detector combinations and enables cross-checks of the results in overlapping energy bands. We present the method, parameters for data selection, and the status of this analysis.