The robust air traffic network relies on safe and punctual arrivals to and departures from airports. Efficient aircraft ground operations and maintenance thus significantly contribute to stable traffic in- and outbound flows. Any improved prediction of the aircraft turnaround time can help reduce local delays and their propagation through the network. Key is forecasting the related operational states to allow for adjusted planning and delay mitigation strategies. In this paper, we target to predict incrementally the turnaround time by means of machine learning classification algorithms based on reallife data collected at the airport. A turnaround sub-processes fusion model for improving the forecast precision is developed to integrate the sequential information from the turnaround pattern, which mainly considers the duration of the various turnaround sub-processes and their overlapping conditions. Results indicate that the data-driven fusion model enhances the robustness and reliability of the aircraft turnaround time prediction. It so can efficiently support airport management. We show, that the presented methodology holds universal character, can be applied to any airport holding a significant demand/capacity ratio.