Research Of Discovering High-value Passengers Of Civil Aviation Based On Hadoop

With the deepening of civil aviation information, a large number of PNR(Passenger Name Record) data accumulated in the booking system of every airline. In the case of the frequent passenger membership program in VIP system cannot form an effective attraction to passengers, to the airlines, it is an urgent question that how to use the PNR data of non-frequent passenger to identify the high-value passengers. This paper mainly researches the method of discovering high-value passengers of civil aviation based on Hadoop and its application, so that providing reference for airlines to discover high-value passengers of civil aviation based on PNR data and make effective strategic decisions.To the massive PNR data sets, the conventional approach is hard to process. This paper put forward to cluster computers, establish Hadoop distributed parallel processing platform,store data by HDFS distributed file system, and process the PNR data sets by Map/Reduce data processing model. The experimental results show that, it is quick and effective to process the massive PNR data sets by Hadoop.Focus on problem that K-Means cluster algorithm suffers much from isolated data object,and easily fall into local optimal solution. This paper put forward to optimize K-Means cluster algorithm, compute the Euclidean distance between all data objects in the data set by Hadoop,and count sum of distance, exclude much isolated data objects, optimize the selection of initial center point. The experimental results show that, optimized K-Means cluster algorithm could exclude much isolated data objects effectively, the clustering results more close to the actual data distribution.Above all, this paper put forward a method of discovering high-value passengers of civil aviation based on Hadoop. At first, process the PNR data sets quickly by Hadoop. Then, put forward the optimized RFD passenger value model, and according to analytic hierarchy process(AHP), quantify the expert experience into the weight of each index. At last, make clustering by optimized K-Means cluster algorithm, so that discover high-value passengers,and analyze the characteristics of customers. The experimental results show that, the method is accurate and effective, could discover high-value passengers of civil aviation quickly.