Reconciling Multiple Categorical Preferences with Double Pareto-Based Aggregation

摘要: Given a set of objects and user preferences, both defined over categorical attributes, the Multiple Categorical Preferences (MCP) problem is to determine that are considered preferable by all users. In naive interpretation MCP, matching degrees between users aggregated into single score which ranks objects. Such an approach, though, obscures blurs individual can be unfair, favoring with precise preferences detailed descriptions. Instead, we propose objective fair MCP problem, based on two Pareto-based aggregations. We introduce efficient approach transformation attribute values index structure. Moreover, extension for controlling number returned An experimental study real synthetic data finds our index-based technique order magnitude faster than baseline scaling up millions