Analysis and Prediction of University Websites Perceptions by Different User Groups

Users' subjective impressions of websites are actively being researched, as they are formed very early in the interaction and affect the whole subsequent user experience. The subjective visual complexity is known to affect both website users' cognitive load and overall affective impressions. In our work we construct artificial neural network model to predict users' perceptions of selected universities' websites without the actual users. To train the network, we joined the two kinds of data that we specially collected. The visual complexity-related metrics of websites were extracted from web interface screenshots by our dedicated visual analyzer software implementing the “human-computer vision” approach. The assessments of the homepages orderliness and overall complexity were collected from 61 human annotators of different ages and cultural/national groups. We were able to achieve moderate subjective perceptions prediction accuracy, with relative error of 73.0%. The analysis of the factors' importance suggests that the proposed index of visual-spatial complexity had considerably higher importance than the baseline frequency-based entropy measure for images. The results of our work can aid in controlling the visual complexity perception in website visitors, ultimately contributing to better web usability.