Variations on a theme: Topic modeling of naturalistic driving data

摘要

This paper introduces Probabilistic Topic Modeling (PTM) as a promising approach to naturalisticdriving data analyses. Naturalistic driving data present an unprecedented opportunity to understanddriver behavior. Novel strategies are needed to achieve a more complete picture of thesedatasets than is provided by the local event-based analytic strategy that currently dominates thefield. PTM is a text analysis method for uncovering word-based themes across documents. In thisapplication, documents were represented by drives and words were created from speed and accelerationdata using Symbolic Aggregate approximation (SAX). A twenty-topic Latent DirichletAllocation (LDA) topic model was developed using words from 10,705 documents (real-worlddrives) by 26 drivers. The resulting LDA model clustered the drives into meaningful topics. Topicmembership probabilities were successfully used as features in subsequent analyses to differentiatebetween healthy drivers and those suffering from Obstructive Sleep Apnea.