Research On Potential High-Value Passengers Based On Trip Purposes Mining

How to identify potential high-value passengers from a large number of passenger booking records is an important issue for airways.The existing passenger value evaluation methods rely on a large number of passenger booking records and can only find the current high-value passengers who have more booking records.Potential high-value passengers couldn't be effectively discovered due to the limited historical booking records.On the other hand,passengers travel based on certain trip purposes.Trip purposes exist objectively and are shared by all passengers.Trip purposes can be obtained through large scale passenger booking records.By discovering the distribution of passenger trip purposes,we proposed the research on potential high-value passengers based on trip purposes mining,which has great research value and application value for airline passenger segmentation and passenger value discovery.Potential high-value passengers are passengers who currently have fewer booking records and have higher potential airline demands.Therefore,a probabilistic model based on trip purposes mining is proposed,in order to optimize passenger potential value calculation.Trip purposes are introduced into the model and parameters are adjusted dynamically to discover potential airline demands of each passenger.Experiments show that the proposed model can identify the potential high-value passengers more accurately than the traditional passenger value evaluation methods based on passengers' cumulative number of flight times,passenger's cumulative mileage and RFM model.Since the probabilistic model is proposed with no account of flight cost,and trip purposes among passengers with co-travel relationship are affected each other.Passenger co-travel relationship is considered to find out the exact distribution of passenger trip purposes.Then the weighted sum of the multiple results between probability of passengers on choosing each airline and the airline distance is the passenger's future cost.Therefore,RFMPro model is proposed to find potential high-value passengers by improving the calculation methods of F and M indicators in the traditional RFM model.Experiments show that the proposed RFMPro model can identify potential high-value passengers more accurately than the traditional RFM model.