We propose a system that optimizes fashion coordination according to the user’s preference using interactive evolutionary computation. This system consists of two optimization phases: The first phase optimizes a color of a selected item by inputting an impression word that express the preferred atmosphere of the fashion. The second phase optimizes the color combinations of the other items based on the acquired color and inputting an impression word in the first phase. User fatigue is reduced by preventing the presentation of multiple similar solution candidates using CIEDE2000 considering the human sight characteristics and the search was efficiently improved by eliminating the search around the low evaluation individuals using the Tabu method. We verified the effectiveness of the optimization method in the first phase by experiments using pseudo-users.