The purpose of this paper is to study the issues raised by the implementation of a tool aimed at protecting air traffic sectors against overload in a large-scale air traffic system, acting on the basis of short term prediction. It shows how to overcome the computational complexity using a decentralised and co-ordinated system composed of a co-ordination level and a control level. The study points on the co-ordination level which decompose the large sector network into several smaller overlapping subnetworks that can be controlled independently. A modified interaction prediction method is developed using a fuzzy model. This model provides the interaction prediction of the control units on the basis of imprecise information and aggregated reasoning in order to decrease the multiple data transfer between the control and co-ordination levels. Time complexity of the fuzzy model inference is also studied, the antagonistic goals of reducing inference time of the fuzzy rule-base and increasing the accuracy of the interaction prediction is commented. Classical fuzzy inference models and rule interpolation techniques are then compared.