Efficient Evaluation of Multiple Preference Queries

Consider multiple users searching for a hotel room, based on size, cost, distance to the beach, etc. Users may have variable preferences expressed by different weights on the attributes of the searched objects. Although individual preference queries can be evaluated by selecting the object in the database with the highest aggregate score, in the case of multiple requests at the same time, a single object cannot be assigned to more than one users. The challenge is to compute a fair 1-1 matching between the queries and a subset of the objects. We model this as a stable-marriage problem and propose an efficient technique for its evaluation. Our algorithm is an iterative process, which finds at each step the query-object pair with the highest score and removes it from the problem. This is done efficiently by maintaining and matching the skyline of the remaining objects with the remaining queries at each step. An experimental evaluation with synthetic and real data confirms the effectiveness of our method.