Detecting Transportation Mode Using Dense Smartphone GPS Trajectories and Transformer Models

TL;DR

This paper introduces SpeedTransformer, a Transformer-based model that accurately detects transportation modes from dense GPS data, outperforming traditional models and demonstrating strong transfer learning capabilities in real-world scenarios.

Contribution

The study presents a novel Transformer model for transportation mode detection using only speed data, with superior performance and transferability compared to existing methods.

Findings

SpeedTransformer outperforms LSTM in benchmark tests.

The model achieves high accuracy across different regions after fine-tuning.

It maintains robustness in complex environments with high data uncertainty.

Abstract

Transportation mode detection is an important topic within GeoAI and transportation research. In this study, we introduce SpeedTransformer, a novel Transformer-based model that relies solely on speed inputs to infer transportation modes from dense smartphone GPS trajectories. In benchmark experiments, SpeedTransformer outperformed traditional deep learning models, such as the Long Short-Term Memory (LSTM) network. Moreover, the model demonstrated strong flexibility in transfer learning, achieving high accuracy across geographical regions after fine-tuning with small datasets. Finally, we deployed the model in a real-world experiment, where it consistently outperformed baseline models under complex built environments and high data uncertainty. These findings suggest that Transformer architectures, when combined with dense GPS trajectories, hold substantial potential for advancing transportation mode detection and broader mobility-related research.