Prediction of the Emotion Responses to Poster Designs based on Graphical Features: A Machine learning-driven Approach

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 167
  • Download : 0
Background Advances in data science have allowed us to investigate human emotions based on the learning of the data set, with current technology enabling us to research the relationship between the form elements of graphic images and human emotional responses. However, to date, studies have been limited to photography, meaning that the practice of graphic design remains unexplored. Methods We used a design workshop to exploit graphic elements that are perceptually relevant. By identifying and quantifying the elements of each image (N = 320), we were able to define the machine learning features. We collected emotion assessments relating to the images with regard to pleasure, arousal, and dominance using a Likert scale, and combined them with the features to build a dataset. The assessments were carried out twice to form a baseline of accuracy. Applying machine learning methods such as kknn, svmRadial, and C5.0, we modeled algorithms that predict individuals’ emotional assessments of specific aspects. Results Human assessments were repeatedly in the range of 62.81–80.93%, and a prediction accuracy between 52.26–80.32% was achieved. In particular, the prediction accuracy of pleasure and dominance aspects was relatively high across all individuals. Conclusions This study demonstrates the use of machine learning-driven algorithm modeling to predict assessments of emotion based on graphical elements of movie posters. Overall, predictions of pleasure and dominance aspects showed higher accuracy than those of the arousal aspect. Limitations and further investigations are discussed to obtain not only a more accurate but also a more insightful estimation.
Issue Date
Article Type

Archives of Design Research, v.33, no.2, pp.39 - 55

Appears in Collection
ID-Journal Papers(저널논문)
Files in This Item
There are no files associated with this item.


  • mendeley


rss_1.0 rss_2.0 atom_1.0