This is a partial implementation for paper LHMM: A Learning Enhanced HMM Model for Cellular Trajectory Map-matching
The following modules are required.
- Ubuntu 16.04
- Python >=3.5 (
Anaconda3
recommended) - PyTorch 0.4 (
virtualenv
recommended) - Cuda 9.0
Due to privacy protocols, we cannot render the Map-matching part. We try our best to provide the interface of the learned observation and transition probabilities. Moreover, a simple HMM framework is also provided in src/pymatch.py
.
$ cd pygcn
$ python gcn_train.py
$ cd trans
$ python trans_train.py
$ python gcnServer_old.py
$ python transServer.py
We provide some figures to illustrate the matching of LHMM in director Mapmatching-Figures
.
We used real trajectory data. Unfortunately, due to privacy protection, we cannot provide the dataset for testing.
The code is developed under the MPL-02.0 license.
If you have any questions or require further clarification, please do not hesitate to send an email to us (E-mail address:shiweijie0311@foxmail.com)