This is the code for the SenSys '20 poster "Localization from activity sensor data".
A Conda *.yml file is provided for recreating the development environment. Just run conda env create -f environment.yml to install all necessary packages.
train_models.py trains the model using auto-sklearn, it is set to take 3 hours
localize.py runs the localization and puts the results into a *.csv file
results.ipynb visualizes the results and stores the plots, we used in the paper
To run, activate the Conda environment, place the data into the data folder and type python train_models.py && python localize.py.
We use proprietary cow data, which we can unfortunately not share. We use MERRA2 data, the *.csv files for each position need the corresponding postal code as prefix, e.g. 1210_SoDa_MERRA2_lat48.283_lon16.412_2019-01-01_2019-12-31_2029583187.csv.
If you find our work useful, please cite it using the following BibTex entry:
@inproceedings{papst2020,
title = {Localization from activity sensor data},
author = {Papst, Franz and Stricker, Naomi and Saukh, Olga},
booktitle = {Proceedings of the 18th Conference on Embedded Networked Sensor Systems},
pages = {703--704},
year = {2020}
}