Vessel Destination and ETA Prediction
A deep learning model predicting vessel destination and ETA for freight market monitoring.
Business issue
Why I was brought into the project
Bulk shipping traders needed better visibility on vessel activity by region, including destination and estimated time of arrival.
Context
The environment around the project
Freight traders needed predictive visibility on vessel destinations and ETA.
Functional environment
The functional context covered freight trading, vessel monitoring, regional forecasting and business indicator tracking.
Technical environment
The technical environment combined R, R Shiny, deep learning, time series, SQL, Jenkins and Git on Bitbucket.
Challenges
The model required data collection from satellites, external databases and web scraping, followed by transformation, prediction and dashboard delivery.
Solution
My contribution and its impact
My contribution to the project
I collected and integrated vessel data, prepared it for modeling, developed a neural network to predict destination and ETA, automated pipelines and built R Shiny dashboards.
- Neural network for vessel destination and ETA prediction
- Automated data pipelines
- R Shiny trader dashboard
- R functions and packages for freight analytics
Impact
The project gave traders a more predictive view of vessel movements and supported market monitoring with automated analytics.
- Improved vessel activity forecasting
- Better regional monitoring for bulk shipping traders
- Automated predictive analytics workflow
Impact metrics
Approach
How I structured the work
- Collect vessel data from satellites, external databases and web scraping.
- Prepare features and train a neural network for destination and ETA prediction.
- Automate pipelines with Jenkins and expose insights through R Shiny dashboards.
Takeaways
What I learned from this project
- Forecasting systems depend heavily on data integration quality.
- Dashboards are more valuable when they expose predictive signals in the trader workflow.
- Automation is necessary when models need frequent updates from external data sources.