Continuous Prediction of Web User Visual Attention on short span Windows based on Gaze Data Analytics

TL;DR

This paper introduces a method to predict web users' visual attention in short time windows by analyzing gaze data, enabling dynamic understanding of user focus without detailed site structure knowledge.

Contribution

It proposes the concept of visit intention and demonstrates its prediction using gaze data and multilabel classification, with high accuracy from limited user data.

Findings

Achieved 84.3% AUC in prediction accuracy

Attained 79% overall accuracy

Identified visual kinetics features as consistently important

Abstract

The existing approaches to identify personalized salience zones of a Web page do not consider the dynamic behavior in time of the Web user's gaze or the alterations of its content. For this reason, this paper proposes the concept of visit intention, an indicator of the visual attention of a Web user in a certain period of time, short span time windows, in different areas of interest. This indicator gives information on the areas of interest of a website that will be visited by a user over a time window, without requiring to know the structure of the site in each window. Our approach leverages the population-level general gaze patterns and the user's visual kinetics. We show experimentally that it is possible to conduct such a prediction through multilabel classification models using a small number of users, obtaining an average area under curve of 84.3 %, an average accuracy of 79 %, and an individual area of interest accuracy of 77 %. Furthermore, the user's visual kinetics features are consistently selected in every set of a cross-validation evaluation.