Causal Feature Selection for Individual Characteristics Prediction

TL;DR

This paper introduces a causal feature selection method to improve the prediction of individual characteristics from behavioral data, effectively handling noise caused by context factors.

Contribution

It proposes a novel causal identification approach for feature selection that enhances predictive accuracy in individual characteristic inference.

Findings

Causal feature selection improves prediction accuracy.

Models trained with causally selected features outperform existing methods.

Demonstrated effectiveness on theme park visitor data.

Abstract

People can be characterized by their demographic information and personality traits. Characterizing people accurately can help predict their preferences, and aid recommendations and advertising. A growing number of studies infer people's characteristics from behavioral data. However, context factors make behavioral data noisy, making these data harder to use for predictive analytics. In this paper, we demonstrate how to employ causal identification on feature selection and how to predict individuals' characteristics based on these selected features. We use visitors' choice data from a large theme park, combined with personality measurements, to investigate the causal relationship between visitors' characteristics and their choices in the park. We demonstrate the benefit of feature selection based on causal identification in a supervised prediction task for individual characteristics. Based on our evaluation, our models that trained with features selected based on causal identification outperformed existing methods.