Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees
The vessel speed is one of the important parameters that govern safety, emergency, and transport planning in the Arctic. While previous studies have traditionally relied on physics-based simulations to predict vessel's speed in ice-covered waters, most have not fully explored data-driven approa...
Published in: | Maritime Transport Research |
---|---|
Main Authors: | , , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Elsevier Science
2021
|
Subjects: | |
Online Access: | https://hdl.handle.net/11250/3029508 https://doi.org/10.1016/j.martra.2021.100024 |
id |
ftntnutrondheimi:oai:ntnuopen.ntnu.no:11250/3029508 |
---|---|
record_format |
openpolar |
spelling |
ftntnutrondheimi:oai:ntnuopen.ntnu.no:11250/3029508 2023-05-15T14:55:47+02:00 Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees Rao, Prithvi S. Kim, Ekaterina Smestad, Bjørnar Brende Asbjørnslett, Bjørn Egil Bhattacharyya, Anirban 2021 application/pdf https://hdl.handle.net/11250/3029508 https://doi.org/10.1016/j.martra.2021.100024 eng eng Elsevier Science Sigma2: NS9672K Maritime Transport Research. 2021, 2, . urn:issn:2666-822X https://hdl.handle.net/11250/3029508 https://doi.org/10.1016/j.martra.2021.100024 cristin:1912741 Navngivelse 4.0 Internasjonal http://creativecommons.org/licenses/by/4.0/deed.no CC-BY 19 2 Maritime Transport Research 100024 Maskinlæring Machine learning Skipsfart Shipping Arktis Arctic VDP::Kunnskapsbaserte systemer: 425 VDP::Knowledge-based systems: 425 Peer reviewed Journal article 2021 ftntnutrondheimi https://doi.org/10.1016/j.martra.2021.100024 2022-11-09T23:42:01Z The vessel speed is one of the important parameters that govern safety, emergency, and transport planning in the Arctic. While previous studies have traditionally relied on physics-based simulations to predict vessel's speed in ice-covered waters, most have not fully explored data-driven approaches and powerful supervised machine learning tools to aid speed prediction. This study offers a perspective of applying supervised machine learning models to predict MV SOG using historical Automatic Identification System (AIS) data and without explicit knowledge of local ice conditions. This paper presents a case-study from the region of the Eastern Barents Sea and the Southern Kara Sea. We first analyzed the vessel traffic situation for the years 2017 and 2018, and then used this knowledge to build statistical models to predict vessel speeds. Finally, we evaluated the models’ performance on a test dataset from January 2019. Performance of three models (Random Forest, XGBoost, and LightGBM) have been tested with a variety of date-time handling techniques, and data input mode being permuted to arrive at the most optimal model. The results demonstrate the ability of the models to predict the vessel's speed based on its geographical location, time of the year and other engineered features such as daylight information and route. With the proposed approach we were able to achieve mean absolute error 3.5 knots in average on a test dataset without explicit knowledge of local ice conditions around the vessel, with the majority of the errors being in the Kara Strait region and the Sabetta Channel. publishedVersion Article in Journal/Newspaper Arctic Arktis Arktis* Barents Sea Kara Sea ice covered waters NTNU Open Archive (Norwegian University of Science and Technology) Arctic Barents Sea Kara Sea Sabetta ENVELOPE(72.039,72.039,71.214,71.214) Sog ENVELOPE(-20.972,-20.972,63.993,63.993) Maritime Transport Research 2 100024 |
institution |
Open Polar |
collection |
NTNU Open Archive (Norwegian University of Science and Technology) |
op_collection_id |
ftntnutrondheimi |
language |
English |
topic |
Maskinlæring Machine learning Skipsfart Shipping Arktis Arctic VDP::Kunnskapsbaserte systemer: 425 VDP::Knowledge-based systems: 425 |
spellingShingle |
Maskinlæring Machine learning Skipsfart Shipping Arktis Arctic VDP::Kunnskapsbaserte systemer: 425 VDP::Knowledge-based systems: 425 Rao, Prithvi S. Kim, Ekaterina Smestad, Bjørnar Brende Asbjørnslett, Bjørn Egil Bhattacharyya, Anirban Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
topic_facet |
Maskinlæring Machine learning Skipsfart Shipping Arktis Arctic VDP::Kunnskapsbaserte systemer: 425 VDP::Knowledge-based systems: 425 |
description |
The vessel speed is one of the important parameters that govern safety, emergency, and transport planning in the Arctic. While previous studies have traditionally relied on physics-based simulations to predict vessel's speed in ice-covered waters, most have not fully explored data-driven approaches and powerful supervised machine learning tools to aid speed prediction. This study offers a perspective of applying supervised machine learning models to predict MV SOG using historical Automatic Identification System (AIS) data and without explicit knowledge of local ice conditions. This paper presents a case-study from the region of the Eastern Barents Sea and the Southern Kara Sea. We first analyzed the vessel traffic situation for the years 2017 and 2018, and then used this knowledge to build statistical models to predict vessel speeds. Finally, we evaluated the models’ performance on a test dataset from January 2019. Performance of three models (Random Forest, XGBoost, and LightGBM) have been tested with a variety of date-time handling techniques, and data input mode being permuted to arrive at the most optimal model. The results demonstrate the ability of the models to predict the vessel's speed based on its geographical location, time of the year and other engineered features such as daylight information and route. With the proposed approach we were able to achieve mean absolute error 3.5 knots in average on a test dataset without explicit knowledge of local ice conditions around the vessel, with the majority of the errors being in the Kara Strait region and the Sabetta Channel. publishedVersion |
format |
Article in Journal/Newspaper |
author |
Rao, Prithvi S. Kim, Ekaterina Smestad, Bjørnar Brende Asbjørnslett, Bjørn Egil Bhattacharyya, Anirban |
author_facet |
Rao, Prithvi S. Kim, Ekaterina Smestad, Bjørnar Brende Asbjørnslett, Bjørn Egil Bhattacharyya, Anirban |
author_sort |
Rao, Prithvi S. |
title |
Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
title_short |
Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
title_full |
Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
title_fullStr |
Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
title_full_unstemmed |
Predicting vessel speed in the Arctic without knowing ice conditions using AIS data and decision trees |
title_sort |
predicting vessel speed in the arctic without knowing ice conditions using ais data and decision trees |
publisher |
Elsevier Science |
publishDate |
2021 |
url |
https://hdl.handle.net/11250/3029508 https://doi.org/10.1016/j.martra.2021.100024 |
long_lat |
ENVELOPE(72.039,72.039,71.214,71.214) ENVELOPE(-20.972,-20.972,63.993,63.993) |
geographic |
Arctic Barents Sea Kara Sea Sabetta Sog |
geographic_facet |
Arctic Barents Sea Kara Sea Sabetta Sog |
genre |
Arctic Arktis Arktis* Barents Sea Kara Sea ice covered waters |
genre_facet |
Arctic Arktis Arktis* Barents Sea Kara Sea ice covered waters |
op_source |
19 2 Maritime Transport Research 100024 |
op_relation |
Sigma2: NS9672K Maritime Transport Research. 2021, 2, . urn:issn:2666-822X https://hdl.handle.net/11250/3029508 https://doi.org/10.1016/j.martra.2021.100024 cristin:1912741 |
op_rights |
Navngivelse 4.0 Internasjonal http://creativecommons.org/licenses/by/4.0/deed.no |
op_rightsnorm |
CC-BY |
op_doi |
https://doi.org/10.1016/j.martra.2021.100024 |
container_title |
Maritime Transport Research |
container_volume |
2 |
container_start_page |
100024 |
_version_ |
1766327793957732352 |