Isolation Forest for Anomaly Detection in Raw Vehicle Sensor Data

Julia Hofmockel, Eric Sax

2018

Abstract

A vehicle generates data describing its condition and the driver’s behavior. Sending data from many vehicles to a backend costs money and therefore needs to be reduced. The limitation to relevant data is inescapable. When using data collected from a vehicle fleet, the normality can be learned and deviations from it identified as abnormal and thus relevant. The idea of learning the normality with the Replicator Neural Network and the Isolation Forest is applied to the identification of anomalies and the reduction of data transfer. It is compared how good the methods are in detecting anomalies and what it means for the traffic between vehicle and backend. It can be shown that the Isolation Forest beats the Replicator Neural Network. When reducing the transfered amount of data to 7%, in average more than 80.63% of the given anomalies are included.

Paper Citation

in Harvard Style

Hofmockel J. and Sax E. (2018). Isolation Forest for Anomaly Detection in Raw Vehicle Sensor Data.In Proceedings of the 4th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS, ISBN 978-989-758-293-6, pages 411-416. DOI: 10.5220/0006758004110416

in Bibtex Style

@conference{vehits18,
author={Julia Hofmockel and Eric Sax},
title={Isolation Forest for Anomaly Detection in Raw Vehicle Sensor Data},
booktitle={Proceedings of the 4th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS,},
year={2018},
pages={411-416},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0006758004110416},
isbn={978-989-758-293-6},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 4th International Conference on Vehicle Technology and Intelligent Transport Systems - Volume 1: VEHITS,
TI - Isolation Forest for Anomaly Detection in Raw Vehicle Sensor Data
SN - 978-989-758-293-6
AU - Hofmockel J.
AU - Sax E.
PY - 2018
SP - 411
EP - 416
DO - 10.5220/0006758004110416