MASSE: A Multi-Agent Based Air Ticketing Service Selection Approach

Online travel agencies (OTAs) rely on the global distribution system (GDS) to provide air ticket query and booking services. Air companies put their tickets onto a number of ticket counters for sale and OTAs provide value-added services to customers relying on air ticketing services provided by ticket counters. When OTAs query these services, the returned ticket information is redundant. Therefore, OTAs should choose some air ticketing services from GDS to query the latest air ticket information. At the same time, the quality (i.e., the prices of tickets provided) of air ticketing services differs because their tickets come from a relatively fixed set of air companies, who have their specific ticket pricing strategies. Since the quality of an air ticketing service is related to the ticket query it processes , to select appropriate services from all air ticketing services in terms of the content of the ticket query is a difficult task, which can not be solved by current service selection approaches. Therefore, we propose a Multi-agent-based Air ticketing Service SElection approach (MASSE). MASSE is based on a multi-agent reinforcement learning so that it can capture the dynamics of quality of air ticketing services. Extensive experiments show MASSE can reduce invalid queries to air ticketing services while retaining the quality of query results.