Investigation of efficiency of manipulation in interactive Tabu Search for optimizing fragrance composition

Fragrance is one of important media types used in our daily life, and it is ideal that using media content suited to each user. However, adjustment of fragrance composition is difficult for most of general users. Interactive Evolutionary Computation (IEC) is known as an efficient method to optimize media contents, and we have already proposed some IECs for optimizing fragrance composition. To enhance the optimization ability of IEC, some previous studies proposed that IEC accepts user's manipulations on solution candidate in IEC. Our previous study proposed Interactive Tabu Search (ITS) for optimizing fragrance composition with the user's manipulation, and a fundamental efficiency of the ITS was investigated. However, the efficiency was not investigated with comparing experiment. Purpose of this study was to investigate the efficiency of the manipulation in the ITS by comparing with ITS without any manipulation. With these two experimental conditions, smelling experiments were conducted. In result of optimizing experiment, relatively one direction changes in intensity of some aroma sources were observed in both conditions. In result of evaluating experiment, maximum mean fitness was observed in the final best solution in the ITS with manipulation. The fitness of the best solutions in both conditions was significantly higher than solution candidate in the initial generation. However, significant difference between the two conditions was not observed.