Multimodal eye-tracking (gaze, HR, GSR) data, for product choice prediction in e-commerce
Description
This dataset contains eye-tracking and physiological data collected during a controlled choice-based experiment in an e-commerce context. It was produced as part of the thesis "An Intelligent System for Visual Attention Analysis Based on Eye Tracking and Artificial Intelligence Focused on UI/UX", developed at ISEP (Instituto Superior de Engenharia do Porto) in collaboration with the GECAD research group (Research Group on Intelligent Engineering and Computing for Advanced Innovation and Development). Thirty participants completed four product selection tasks, each involving a 2×4 grid of eight products, while gaze behaviour and physiological signals (heart rate, galvanic skin response, pupil diameter) were recorded using a Gazepoint GP3 HD eye-tracker. The dataset is organised into raw exported files (per participant) and a processed, machine-learning-ready CSV file. It supports research in visual attention, consumer behaviour, and predictive modelling of product choice from multimodal signals.
Files
Steps to reproduce
1. Collect gaze and physiological data using Gazepoint GP3 HD eye-tracker with HR and GSR sensors during the choice-based experiment. 2. Define AOIs for each product in each frame using the naming convention FrameX_ProductX_Product in Gazepoint Analysis. 3. Export All_Gaze, Data_Summary, and Fixations files from Gazepoint Analysis software for each participant. 4. Apply the provided processing pipeline to extract and aggregate features at the participant–choice–product level. 5. Handle missing values: set to 0 for all features except TTFF, for TTFF, flag missing observations (TTFV_Missing = 1) and impute with the global median of non-missing values. The resulting dataset contains 960 rows and 25 columns as described in README.md.
Institutions
- Polytechnic Institute of PortoPorto, Porto