Comparative Study of Data Mining Methods for Aerodynamic Multiobjective Optimizations

Practical aerodynamic design problems are typically multiobjective design optimization problems that have multiple contradicting objectives and many design parameters. Goal of multiobjective design optimization is to find Pareto-optimal solutions to reveal tradeoff information between the objectives and effect of each design parameters. Recently, idea of "Multi-Objective Design Exploration (MODE)" was proposed by Obayashi et al. as an approach to find such design information. They proposed to use multiobjective evolutionary algorithm to find Pareto-optimal solutions and to use data mining methods such as Self-Organizing Map (SOM) to extract design information from the Paretooptimal solutions. However, it has not been discussed yet which data mining method is suitable for analysis of Pareto-optimal solutions among many data mining methods.