Family Travels Forecast And Analysis In Passenger Social Networks

With the development of tourism and the rising of residents’ tourism consumption, travel by plane has become a common way. At the same time, the information level of the civil aviation has been improved greatly. A large number of passengers’ personal information and their historical information have not been used efficiently. It is becoming one of the major issues in the field of civil aviation that how to mine more valuable knowledge from the historical operational data. The family is the basic unit of society. Family passenger groups are one of the most popular travel consumption units in the passenger transportation market. Because the family travel is affected by many factors such as time, economic, so the family travel presents uncertainty. If we can make an accurate forecast that if a family will travel over the next period of time, it will help the passenger transportation, tourism and other related industries to provide personalized products or services for families. In turn, it also helps improve service quality, reduce marketing costs and improve family passenger satisfaction.In this paper, we study and define what family travel is, and regard family travel prediction as a classification problem. First of all, according to the personal information and family history of travel information to build the demographic characteristics of the family, history, behavior characteristics and predict the time window, the nature of these characteristics describe the different families and behavioral attributes. First of all, we construct the demographic characteristics, historical behavior characteristics and predicting time window characteristics according to the personal information and the histories of family travel information. These characteristics describe the families’ natural attributes and behavioral attributes. In order to achieve better classification effect, we also construct passenger social networks by extracting co-travel relations between passengers from their historical travel records and generate the social network features of family members to further describe the family characteristics. We use a variety of traditional classification algorithms to predict family travels. Finally, we analyze the travel behavior characteristics on the family groups.The experimental data originates from the real data sets of civil aviation. We compare the effect of various classifiers and find that random forest algorithm is the best classification for family travels forecast of social networks in civil aviation. The experiment result indicates that random forest can forecast whether a family will travel or not in the next month. And the accuracy can reach above 85%. It also provides support for the decision of the passenger transportation market. The statistics and analysis of family travel behaviors also provides support for the decision of the relevant organizations’policy makers.