Abstract: In order to improve the accuracy of prediction of risks in flight operations, Shandong airlines flight control workflow was analyzed first, and fifteen risk items were taken initially as flight operation risk assessment indicators according to the Civil Aviation Authority Advisory Circular "Air Carrier Operation Control Risk Management System Implementation Guide". Then, on the basis of the data on 100 historical flights of Shandong airlines, the number of risk items was reduced to eight by using the rough set theory, the genetic algorithm and the Johnson's algorithm. Finally, a risk prediction model was built by means of the SVM algorithm. A simulation was carried out with Matlab. The results show that the overall correct rate of sample classification can reach 82.22% for high, medium, low-three types of risk level, and that the method can be used to realize the assessment and classification of risks in flight operations.

Key words: flight operations, prediction of risk, rough sets, genetic algorithm, Johnson's algorithm, support vector machine(SVM)