Predicting vessel arrival times in winter conditions : A data-driven approach
Vuotila, Risto (2024)
2024
Tietojohtamisen DI-ohjelma - Master's Programme in Information and Knowledge Management
Johtamisen ja talouden tiedekunta - Faculty of Management and Business
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2024-12-20
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-2024112210411
https://urn.fi/URN:NBN:fi:tuni-2024112210411
Tiivistelmä
Maritime transportation covers 80% of the volatility of the imports in Finland. At the same time, the Baltic Sea is one of the areas where the sea does freeze during winter. The ice complicates maritime navigation by affecting vessel speed and causing them to be beset. Inaccurate estimations of arrival times cause inefficiency for example in ports.
The research took a data-driven approach to find out how the vessel movement during winter could be predicted by utilizing machine learning models. The target was to identify which aspects affect vessel movement and to rank them based on their significance.
The study combines theory from two fields: data analysis and maritime transportation. The research describes how to use data analysis in different kinds of projects. Based on the theory, the CRISP-DM data analysis framework was selected and modified to the needs of this research. We explored some commonly used machine learning models and used a real-life use case to validate the usability and performance. The data that was used in the analysis included ice, wind and vessel information data and was mainly collected from public APIs.
The study answers several topics. In this research, we introduce a workflow on how dataoriented research could be performed. We describe how to build an environment which enables an agile way of performing iterative analysis. The study estimates the importance of different environmental variables and how they should be taken into consideration when estimating the vessel arrival time. We provide a comparison of different machine learning models which could be utilized when answering traffic-related problems. Also, the study describes the preprocessing steps which are required to be performed before the analysis itself.
The research took a data-driven approach to find out how the vessel movement during winter could be predicted by utilizing machine learning models. The target was to identify which aspects affect vessel movement and to rank them based on their significance.
The study combines theory from two fields: data analysis and maritime transportation. The research describes how to use data analysis in different kinds of projects. Based on the theory, the CRISP-DM data analysis framework was selected and modified to the needs of this research. We explored some commonly used machine learning models and used a real-life use case to validate the usability and performance. The data that was used in the analysis included ice, wind and vessel information data and was mainly collected from public APIs.
The study answers several topics. In this research, we introduce a workflow on how dataoriented research could be performed. We describe how to build an environment which enables an agile way of performing iterative analysis. The study estimates the importance of different environmental variables and how they should be taken into consideration when estimating the vessel arrival time. We provide a comparison of different machine learning models which could be utilized when answering traffic-related problems. Also, the study describes the preprocessing steps which are required to be performed before the analysis itself.