Multinomial thresholded LASSO for interpretable dimension reduction of human activity sequences

TL;DR

This paper introduces a thresholded LASSO approach for reducing dimensionality in human activity sequences, effectively identifying key positions that distinguish different behavior patterns from high-dimensional categorical data.

Contribution

It develops and evaluates a novel regularization method tailored for sequence data, demonstrating its superiority over traditional techniques in capturing complex temporal dependencies.

Findings

Thresholded LASSO outperforms established methods in sequence data analysis.

The approach effectively identifies key sequence positions for behavior differentiation.

Method shows promise for interpretable dimension reduction in high-dimensional categorical sequences.

Abstract

The widespread collection of data from mobile and wearable devices has created unprecedented opportunities to study human behavior in fine temporal resolution. One common structure for such data is categorical sequences: ordered, multinomial observations across many time points. These sequences present unique statistical challenges due to their high dimensionality and complex temporal dependence, including both short- and long-term correlations. Yet, there has been relatively little methodological development focusing on principled dimension reduction specifically tailored to this type of data. In this paper, we develop and evaluate approaches to identifying "key" sequence positions which distinguish sequence types. We frame this challenge as a regression problem, introduce a variety of regularization techniques that could be applied to achieve position-based dimension reduction, and evaluate them on the motivating dataset that reflects daily time use patterns collected via a smartphone application. Results show that the thresholded LASSO, a relatively underused technique, performs better than more established methods for data with complex sequential structure.