In the case of solving the problem of that the evaluation index of frequent flyer program is single, cannot identify high-value passengers accurately, present a method of discovering high-value passengers combines Map/Reduce and data mining. Processing gigabytes of PNR data on Hadoop by Map/Reduce parallelly, according to the improved RFD model and analytic hierarchy process, determine the customer value indexes and the weight of each index, identify the high-value passengers by data mining, and make an experiment on a real PNR dataset. The experimental result shows that, the method can effectively identify the high-value passengers of airline and provide a favorable basis for airlines to make effective decisions.