Addressing the Cold-Start Problem in Personalized Flight Ticket Recommendation

With the rapid development of the tourist industry, an increasing number of passengers book flight tickets through online travel agencies. When searching for a flight ticket online, users are overwhelmed by the choice on offer. Even though flight ticket recommendation has been widely investigated, the current approaches are unable to recommend flight tickets that meet an individual's preference efficiently because of the severe cold-start problem. This paper provides strategies to address the cold-start problem for flight ticket recommendation. We conduct an exploratory study over the real-world flight ticket recommendation scenario and classify the cold-start problem of flight ticket recommendation into two categories, namely route cold-start and user cold-start. We propose methods based on route similarity and social relationships between passengers to improve user models. Finally, we map an enhanced user preference model and flight features to latent factor spaces to generate the recommendation results. The experimental results on a real=world data set demonstrate the effectiveness of the proposed approaches.