Air Cargo makes up an important part of the worldwide global trade, measured by volume air cargo makes up 35 percent of the global freight. Due to the fact that shipments can be canceled on short notice or not show up without announcement with little repercussions for the shipper, revenue management is a challenging task for cargo carriers. To counter this risk of capacity underutilization and therefore lost revenue, more shipments are routinely booked on a flight than there is capacity for. This practice is called overbooking and requires the establishing of overbooking levels that depend on the expected show up rate of the booked shipments. A previous master thesis built a random forest model based on a framework proposed by Becker (2011). The model estimates the probabilities for no-shows and cancellations based on historical records of updates for bookings. This thesis focuses on the implementation of changes to improve the performance of the model and on the evaluation of the model regarding important features and the performance over time.