Abstract: With the rapid development of Internet and mobile terminal technology, more and more users prefer to query and book airline tickets on the Internet. In order to relive the heavy operating pressure of online ticketing systems brought by a large number of users’ queries, caching the numbers and prices of tickets (i.e. tickets inventory) has become a widely adopted mechanism by most online ticketing systems. The key issue of cache systems is how to set the time-to-live (TTL) of each query keyword. This paper proposes a cache optimization strategy based on the analysis of users’ query behaviors, which mines the inventory change patterns based on massive amounts of user query logs, and dynamically sets the TTL by forecasting inventory change time interval. We carry out experiments in a real user query dataset collected from an online ticketing site, and the experimental results demonstrate that the proposed method can greatly improve the hitting ratio of cache while ensuring the accuracy of the query results.

Key words:  online ticketing system, user behavior analysis, inventory change, time to live