Introduction

So today, I would like to explain about our ODANA's maritime intelligence engine and automated port performance assessment and anomaly detection system, which is based on expertise combined with machine learning infrastructure. infrastructure.

Why Maritime Intelligence Matters

Eighty percent of the global trade moves by sea, so we can say that it's the backbone of the world trade.

But at the same time, with this dependency, it's also a critical point of failure, disrupts global operation schedules. So it's a very critical critical point.

Critical Events and Real-World Context

I share straight off HURMUS data, AS data, as an example. So I will try to explain also applications which is related with this type of critical times.

Current Challenges with Port and AIS Data

So, with the current system, we have general three main problems. One of them is the, especially port -analyzed data, official data, manually implemented and official data are published after one month. So this creates especially related with port performances delay about the decisions and other implementation.

Second again labor intensive work to implement these datas and mostly done by manual work.

And third 1Third problem is main problem with this data implementation of AIS data is they need to be tailored for different type of geographic situations and areas. areas.

ODANA’s Unsupervised Maritime Intelligence Engine

So our answer is appropriate unsupervised analysis engine, which is a product of a PhD thesis in Naval Defense Academy in Japan.

and with this thesis we now implementing in industry so we started to implement the product for the industry use and the main idea is to use this data to have two outcomes one of them is port efficiency the second one is anomaly detection.

Key Technical Differentiators (Trajectory-Based Features)

So again with what makes us different with the present systems first to implement data instead of 12 data features we are just using six trajectory features about AIS data and instead of point -by -point implementing we are using trajectory segments of this data and again based on our PhD thesis we have that we compare the

Accuracy and Benchmarking Against Official Data

outcome of our system our algorithm with official data we have 93 percent of accuracy and compared Compared to global ports, again with the official data, we have reached again 98 -99 % of classification accuracy. So these are the differences that we are implementing with the algorithm.

Application 1: Real-Time Port Performance Intelligence

So there are two applications, possible applications. First port performance intelligence. I will skip those items quickly because we have the data also here.

Automatic Port Boundaries, Zones, and Operational Metrics

1And the engine, our engine identifies the port boundaries, berthing locations, anchorage zones, and board locations directly from ship behavior and from their patterns. so that we can calculate port around time, turnaround time, waiting time, and connectivity metrics, which will be helpful for the port management and also logistic companies, forwarding companies.

Real-Time Delivery vs. One-Month Reporting Delays

So we are just doing these calculations in real time. UN trade and development organization is publishing those type of information after one month of delay. So we are supplying with algorithm within some seconds, I can say, with the real -time data.

When we compare with official data, we just have 2 .9 % of mean error between three main validation ports, which are Indonesia, Malaysia and Singapore ports. So these are, when we compare with the official UN data, we just have 2 .9 % of error rate. But instead of waiting months, we just give on time.

Application 2: Anomaly Detection for Operations and Security

So the second application is anomaly detection. so with this information so we can real -time understand about deviation unexpected activity port arrival times possible delays and other stuff close

approach zone entry and loitering especially loitering is important in these days it's just a sign of some port sort of illegal operation or to be be followed so these are typical outcomes of our algorithm and software

Validation Results and Detected Anomalies

so with this validation process we have just tested 209 vessels and our engine in successfully detecting nine known anomalies beside five additional anomalies which is not detected by experts so algorithm shows better performance so again the one of the critical

Why Loitering Detection Matters (Peace and Wartime)

detection anomaly detection is loitering so that this especially during war times like in these days so loitering is also critical importance about mine sweeping and mine or smuggling or any kind of illegal activities so it's also an important real -time data for the decision makers so the other advantage that our

Geography-Free Deployment Across Different Port Styles

algorithm is giving it's it has total geography free algorithm so you don't have to fine -tune for different locations so this is the example of four different port styles in different parts of the world so that we can obtain the same result without making any private fine -tune separate fine -tune and the

algorithm can be deployed same day without any delay or without any training with the system

Opportunities and Target Use Cases

so what are the opportunities for us is the accuracy is 99 percent compared to official data and we have 2 .9 percent error rate but on time instead of waiting months and thanks to the global thanks to the uniform and single algorithm which is suitable for all the way it's also no need for regional training and last it

Commercial Supply Chain and Governmental Security Applications

will be applicable for commercial supply chain efficiency and also governmental activities so So I can summarize the application in two parts, civil use for the decision maker for the logistic, maritime

logistic, and also governmental security points. So that's all about the general presentation about the product.

Demo Walkthrough (Singapore Region Example)

And I have a short video about a couple of minutes instead of demo.

I hope I have on time so this is AS data is just a bunch of numbers so if you are aware of some money time hobby you can also see about AIS data and how is

Port Status, Turnaround Times, and Anchorage/Security Views

implemented so in this part we just want to show the conditions around Singapore port and close proximity of Malaysia Indonesia and also Singapore so that in this case with 400 ships data which is available in this this demo we just show the capabilities of port anomaly detection and port status data so if you

If you consider Singapore port, it's the biggest, one of the biggest ports in that region, and so by just using AS data online, you can have all the port turnaround times, and also for

security reasons, the anchorage area, as you can see, the round ones which are in anchored ships, so that you can see if there is a ship outside the boundaries, so the port security can identify about this type of security problems, so that they can manage on time.

Insurance Use Cases, Loitering Examples, and Common Trajectories

So it's also important for decision makers like insurance companies to have the calculation about their claims and problems in case of any payment requirement so this helps also as a tool for official tool to prove this is another example of

loitering again in the area so you can see the ships which are not in correct packed and just pet and just making some loitering so this could be normal but also it's a sign of some kind of anomaly in peace and war times so that it's also

helpful tool and the last function as you can see is the common trajectories of the port. This is also helpful to understand about security concerns and also good for again insurance companies' decision makings.

Conclusion and Q&A

So that's it for the product.

Thank you.

And we are here with the team also here. We came from Turkey and we are ready to answer to questions at the end of the thank you thank you