Use Case Talks Series: Challenges in Time Series Prediction
We enjoy in-depth technical discussions and exchange information about interesting technical challenges among industry, academic and individual members in the Use Case Talks Series of the alliance! In a relaxed atmosphere with about 15 – 30 members of the alliance we will brainstorm about possible pathways towards solutions for the presented challenges. Each speaker presents his or her use case in 25 minutes followed by discussion of about 20 minutes.
The next use case talk will take place
20.04.2018, 17:00 – 20:00
at the premises of our host AWK Group. Please find a map here.
A small apero will be kindly provided by our host and member AWK.
Please register by providing your Name, Email, and Organization by email to firstname.lastname@example.org by Wednesday, 18.04.2018
We will feature two use case talks:
Use Case 1
Dr. Michel Philipp
Advanced Analytics Solutions for the NOVA Platform
SBB-IT is currently developing a new standardized platform called NOVA (Neue ÖV Anbindung) that provides distribution services to ticket channels of all transportation providers within the Swiss public transport network. In this talk/workshop, we will discuss our ideas for advanced analytics solutions for improved monitoring, alerting and error handling on the NOVA platform.
Use Case 2
Dr. Peter Kauf, Prognosix
Ideas on how to predict rare event time series with potential use of aggregate information.
Prognosix is currently working in various prediction projects, where we are successfully using machine learning methods on e.g. demand in distribution centers, regional harvests, regional road accident risks. But how do we predict time series at a single site, e.g. demand in a small store, harvest at a single producer’s farm or accident risks on a 2km piece of road? These sites are typically exposed to a high amount of randomness, but also to relevant systematic effects. In this use case talk, we would like to discuss approaches and ideas for separating these systematic effects for single site forecasts, potentially making use of aggregate information.