Abstract: User preferences can influence the user choices in many cases, and the question of preference query becomes an increasingly important issue in relational databases. In many applications, qualitative preferences can be applied more widely than quantitative preferences. In the existing studies of multi-attribute preferences, preference attributes do not have a dependency relationship, but CP-nets(conditional preference networks) is a graph model that represents multi-attribute qualitative preferences with dependencies. At present, the processing of preference queries mainly uses dominance queries and compares the two outcomes one by one to obtain the outcome that satisfies the user preferences. It can be found that comparing the outcomes in pairs causes great waste of resources. Reduce the number of comparisons of its outcomes is examined. The pruning technique is proposed to be applied to dominance queries, and the path of the flipping sequence is pruned so as to effectively reduce the space for database search.

Key words: conditional preference networks(CP-nets), the induced graph of CP-nets, flipping sequence, suffix fixing, least-variable flipping, forward pruning