Webpage Layout Evaluating System Guided By Evolutionary Aesthetics

Evolutionary aesthetics makes it possible to study aesthetics using scientific methods introduced from psychology,statistics and information theory.This research conducted empirical experiments to prove certain principles proposed by evolutionary aesthetics.Guided by these utilizable principles,we targetedly extracted image features to achieve a webpage layout evaluating system with the help of machine learning methods.Fluency theory based on evolutionary aesthetics proposed that observer's aesthetics response is positively correlated to the fluency of his/her eye movement.This study tracked 30 observers' initial landings for 40 web pages(each displayed for 3 seconds)and introduced Shannon Entropy to analyze the data.The result shows that the heatmap entropy(visual attention entropy)is highly correlated with the observers' aesthetic judgements of the web pages.Its improved version,the relative Visual Attention Entropy,has a more significant correlation with the perceived aesthetics(r=-0.65,P<0.0001).This single metric along can distinguish between good-and bad-looking pages with an accuracy of 87.5%.Further investigating reveals that the performance of Visual Attention Entropy became stable after 1 second of exposure.The outcome indicated that the performance could be better if the tracking time was extended beyond 3 seconds.After the prove of the correlation between eye movement fluency and perceived aesthetics,this study dicusses graphic features' predictive ability to aesthetic ratings.We collected a total amount of 1447 web pages and rated them by conducting an online survey.By extracting graphic features that are related to visual complexity and visual saliency,we found grid complexity,margin distribution,information density and saliency distribution performed significant variance between good-and bad-looking groups.Finally we realized an application for the aforementioned results,a web layout evaluating system,and reached a 83%average cross validation accuracy.