A carregar...
Publicação
Sistema inteligente para análise de atenção visual baseado em eye tracking e inteligência artificial focado em UI/UX
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 6.44 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
The rapid growth of e-commerce has intensified the need for personalised and engaging user
experiences. Understanding users’ visual attention and emotional responses while interacting
with digital interfaces is essential for designing adaptive, effective, and user-centered
websites. However, current approaches often address visual attention, emotional analysis,
or interface adaptation separately, lacking a unified intelligent system that integrates these
components.
A systematic literature review following the Preferred Reporting Items for Systematic Reviews
and Meta-Analyses (PRISMA) methodology was conducted to identify existing research
on attention metrics, interface design, emotion recognition, and Artificial Intelligence
(AI)-driven adaptation. The review revealed a gap that this thesis explores, as few studies
link gaze patterns to explicit user preference in a way that could enable adaptive interface
personalisation.
The proposed solution is built around two complementary eye-tracking experiments and a
multimodal predictive pipeline. The first experiment investigated how spatial position and
presentation style influence visual attention within structured product grids. The second,
choice-based experiment required participants to select their preferred product from competing
alternatives, producing a labelled dataset that combines behavioural, oculomotor,
physiological, and contextual features.
Five predictive models were trained and evaluated on the resulting dataset, Random Forest,
Extreme Gradient Boost Classifier, Support Vector Machine, Multilayer Perceptron, and Extreme
Gradient Boost Ranker. Among these, Random Forest achieved the highest overall
performance, correctly identifying the selected product as the top-ranked item in approximately
78% of choice tasks, while the Extreme Gradient Boost Ranker also demonstrated
strong ranking performance. These findings demonstrate that eye-tracking data, combined
with machine learning, can accurately infer user preferences, providing a foundation for
adaptive and personalised e-commerce interfaces.
Descrição
Palavras-chave
Eye-tracking AI Machine Learning Visual attention Product Preference Rank ing UX/UI E-commerce Adaptive Interfaces