Graphic design based on user tags and multi-objective collaborative optimization

Abstract

In graphic design, designers often need to consider multiple goals, such as the esthetics of the design, the accuracy of conveying information, and the visual experience of the user. Multi-objective collaborative optimization can help designers find the best balance point between multiple objectives and improve the quality and effectiveness of their designs. Therefore, research proposed a graphic design method that combines user label preference information with multi-objective optimization, which combines user preference information with PSO (particle swarm optimization) algorithm for multi-objective optimization. The findings showed that when the number of training samples was small, the contour coefficients of each model changed significantly, and the clustering effect of the models was unstable. As the amount of training samples increased, the clustering effect of each model tended to stabilize. The contour coefficients based on improved K-means, K-means++, and classical K-means algorithms were 0.743, 0.707, and 0.546, respectively, indicating that the improved K-means algorithm had the best clustering effect in this study. As the number of users increases, the user accuracy of the research method improves to 0.88. The recall rate of the research method is consistently higher than other comparative methods, at approximately 0.93. The proposed methods in graphic design can effectively integrate user preference information, meet the needs of graphic design, and provide reference for designers in graphic design.