Generalizable User-Specific Mobility Representation Learning Using Autoencoders
Open access
Author
Date
2024-01-12Type
- Student Paper
ETH Bibliography
yes
Altmetrics
Abstract
Addressing the demand for sustainable travel solutions, Axon Vibe leverages a smartphone-based platform to predict commuters’ travel patterns and encourage eco-friendly alternatives. Understanding the pivotal factors shaping individual travel trajectories stands as a key pursuit for the company.
This project focuses on extracting these critical factors and reconstructing travel patterns by employing an Autoencoder (AE) architecture, integrating long short-term memory (LSTM) layers. To enhance mode detection within the latent space, the architecture incorporates a classification framework. The evaluation using t-distributed stochastic neighbor embedding (t-SNE) confirms that the integration of classification improves the intuitive clustering of transport modes within the latent space. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000660086Publication status
publishedPublisher
ETH ZurichSubject
Autoencoders; Travel behavior; Long Short Term Memory (LSTM)Organisational unit
03901 - Raubal, Martin / Raubal, Martin
More
Show all metadata
ETH Bibliography
yes
Altmetrics