What if AI could help rescue at sea?

As part of a hackathon, Avisia has developed artificial intelligence models intended to improve rescue operations at sea by detecting false alerts and allocating resources.

Allassane Mamane, practice manager at Avisia

The potential of artificial intelligence does not only lie in creating value for companies. These technologies can also contribute in other areas, including on a societal level. These purposes can generate meaning for data scientists and thus constitute attractiveness factors for companies developing such use cases.

Read also: With AI, the Autend project is preparing profound changes for the nuclear industry

Consulting firm, specializing in particular in Data, Avisia has designed AI models for rescue operations at sea. This application of artificial intelligence was designed on the occasion of a global hackathon organized by the SAS publisher.

Optimize surveillance and rescue operations

This meeting presented a double opportunity for the firm’s teams. The hackathon thus made it possible to bring out a project imagined internally, but also to test a new technology with the SAS Viya platform in order to promote the development of skills.

To design an AI model, Avisia had previously identified the availability in open data of data from CROSS, the regional operational centers for surveillance and rescue.

These data were integrated into the platform in order to initiate an exploration phase to select the relevant issues. “We started from a principle: how, by using open data and advanced analytics tools, to optimize CROSS operations?”, explains Allassane Mamane, practice manager at AVISIA.

The diagnostic analysis then made it possible to highlight two main subjects: the management of false alarms and the reduction of the number of injuries. The participants of the challenge could then devote themselves to the modeling stage, but after the preparation of data.

The experts also decided to enrich the dataset, in particular by adding meteorological data. For modeling, Avisia mobilized an in-memory processing tool, VDMML, which provides multiple machine learning models. The objective: to test and compare different models to identify the “champion model”.

Machine learning to detect false alerts

On both themes, Avisia has therefore designed its ML models (Gradient Boosting and neural networks), however devoting more time to the subject of “early detection of false alerts”, which potentially would allow CROSS to “gain in efficiency and to make savings in terms of resources.”

The project does not stop at the hackthon, despite a 2nd place in the ranking won in the public sector category. The objective for the firm is now to capitalize on these achievements to design more advanced models in order to offer them to sea rescue operations.

“We have transformed the hackathon into the first stage of an internal project, which we want to see through to completion by proposing an operational solution for CROSS,” says Allassane Mamane.

Avisia has already identified areas for improvement for its models, including enrichment with other data sources. This enrichment is presented “as one of the most important phases”. Among the potential sources, the CROSS could for example have relevant datasets or provide more substantial histories.

Beyond algorithms, the firm is also working on the restitution of model results and their use in the decision-making process. During the hackthon, Avisia used a dashboard in Visual Analytics and Model Studio for the restitution.

Impact and public interest projects for data scientists

As part of a possible agreement with CROSS, the company is considering, for example, ways of making the results of the models available “in near real time” on operational terminals. “The ideal is an incoming alert qualified by the model, which will indicate a probability rate of false alert”, details the Practice Manager.

It is then up to the human to make the final decision. The model aims to help decision-making, not to replace it, especially since the actions, triggered or not, have a potential impact on lives.

Depending on the probability generated by the model, the CROSS could nevertheless allocate appropriate means or trigger proportionate actions. Allassane Mamane is already showing the interest expressed within the Ministry of the Sea.

He further suggests that topics of public interest could be a way to increase the use of AI in the public sector. These technologies stay there still under-exploitedunderlined a report by the Council of State at the end of August.

These projects of general interest are also of interest within specialized companies and among experts in data science. “This is a request from our consultants, in particular to enable them to increase their skills by scanning the entire analytics cycle. We also wanted to have an impact beyond the usual missions that we carry out”, concludes Allassane Mamane.

We would love to say thanks to the author of this write-up for this incredible content

What if AI could help rescue at sea?

You can view our social media pages here and other related pages herehttps://www.ai-magazine.com/related-pages/