How to predict choice using eye-movements data?

Area of Science:

• Consumer Behavior
• Human-Computer Interaction
• Data Science

Background:

• Eye-movement detection technology has advanced, making eye-trackers accessible for research and as computer peripherals.
• This technological progress offers new avenues for measuring participant eye movements in various contexts.

Purpose of the Study:

• To develop and evaluate classification models for predicting food choices using eye-tracking data.
• To identify the most effective classification model for food choice prediction.

Main Methods:

• 112 participants completed 16 choice tasks across four different choice sets (2-, 4-, 6-, and 8-alternative forced-choice paradigms).
• Eye-tracking data was captured using a Tobii X2-60 eye-tracker and Tobii Studio software.
• Thirteen classification models were tested, with performance evaluated using eight parameters and the sum of ranking differences algorithm.

Main Results:

• Decision tree-based techniques consistently outperformed other models across all choice tasks and food categories.
• Quinlan's C4.5 and cost-sensitive decision trees demonstrated the highest performance among the tested classifiers.

Conclusions:

• Decision tree models, especially C4.5 and cost-sensitive variants, are highly effective for predicting food choices based on eye-tracking data.
• Future research should explore fine-tuning these models and their application with mobile eye-tracking devices.