Investigating Progress of Ratios of Delicious Blended Juices in Search by Interactive Evolutionary Computation

抄録

Creating media content suited to each user's preferences and feelings is important. This study treats the search method for delicious blended juices using Interactive Evolutionary Computation (IEC) and especially aims to investigate the proposed IEC's behavior in search space. The investigations on the progress of the ratio contribute to establishing a more practical search method for blended juices. As the concrete algorithm, Genetic Algorithm (GA) was used. Moreover, six source juices composed of three fruit and three vegetable-like juices were employed in the experimental settings as a simpler setting compared with a previous study that could not show an increase in the fitness values in the evaluation process. As the experimental results, gradual increases in mean and maximum fitness values in the search experiment and a marginal increase in the re-evaluation experiment were confirmed. The progress of the convergence in the search area was also clarified in the search. Additionally, higher ratios in fruit source juices and lower ratios in vegetable source juices were observed as common trends in the composition of the best GA individuals. These results are dedicated to establishing a future IEC to find delicious and nutritious blended juices for each user.