Making predictions using incomplete data – a case study of a battery life prediction
In real business problems data is often incomplete. Even when relevant data is being collected, it might not have been collected for long enough to lay the foundation for well-performing models and algorithms. This presentation is about an approach to a specific real-world problem in which a majority of the data was incomplete: building a battery life prediction for batteries with a long longevity using survival analysis. The idea of this presentation is that everyone should learn something new – survival analysis is not very common as a machine learning method, but it can help solve fairly common business problems. We welcome people of all (OK, most) knowledge levels of data analysis and machine learning.
Jonas Mortin works as a data scientist at Verisure Innovation where he is one of the driving forces in empowering Verisure’s systems with AI and machine learning. Previously, he has worked as a business intelligence consultant and a research scientist. He’s always been interested in conveying complex things in a simple manner. Jonas has a PhD in Atmospheric Sciences and Oceanography at Stockholm University, and a Masters in physics at Lund University.