Eingabe löschen

Kopfbereich

Hauptnavigation

Life Sciences und
Facility Management

Research Group for Predictive Analytics

Introduction

«It's hard to make predictions, especially about the future.»

(attr. to Niels Bohr)

Everybody wants to know «what will happen next». We book holidays hoping for fine weather, invest in stocks hoping for good returns, drive another 100km without refuelling expecting that what is left in the tank is still enough to reach the destination. We all engage in speculations and guessing future outcomes, and if we are smart enough, we use additional information to increase our chances of guessing right. And our brains evolved to be exceedingly good at it. But there are limits: we are 3-dimensional creatures, and reasoning in a space exceeding three dimensions is painstakingly hard.

Enter the computer. With the computing power harnessed in silicone, it is now possible to crunch through massive amounts of data, infer missing information, select and extract predictive features and build sophisticated models. It is possible to reason not only in three, but in hundreds and thousands of dimensions, and to discover connections and dependence between variables which could otherwise stay obscure and unnoticed. Equipped with this wealth of knowledge is possible to make predictions – also about the future. Enter the realm of predictive analytics.

Expertise and partners

We bring together the expertise and experience in:

Together with our industrial partners, we strive to solve exciting applied data science problems in the domains of:

SRF Einstein report on the project: Non-invasive wearable core body temperature sensor

Report starting at minute 23.00

Team

Non-invasive wearable core body temperature sensor KTI No. 25392.1 PFLS-LS

Contact: Dr. Krzysztof Kryszczuk

In this project we will develop an innovative, non-invasive, wearable core body temperature sensor. The sensor unit will contain integrated skin temperature and heat flux sensors. The core body temperature will be estimated using state of the art machine learning techniques, incorporating optionally heart rate and accelerometer signals. A demonstrator will be built using the open-source Pebble platform.

Main financing partner: Innosuisse
Project partner: Inselspital Bern, GreenTEG AG

Dynamic Personalized Recommendation System for Hotel Booking Platform KTI No. 19319_2 PFES-ES

Contact: Dr. Krzysztof Kryszczuk

This project aims at the development of a novel personalization system for an online hotel booking platform, consisting of an innovative recommendation algorithm integrated into a big data computational environment. The envisioned deliverable will provide personalized website content recommendations in real time during user's interaction with the booking platform, with an awareness of the dynamics of the user's behavior and intentions.

Main financing partner: Innosuisse
Project partner: UCOB Ventures AG, 2PVentures

Predictive-prescriptive analytics for combustion monitoring in gas turbine power plants

Contact: Dr. Krzysztof Kryszczuk

The project will create a novel software system and service for predictive-prescriptive maintenance of gas turbines. The service will be based on predictive models created from historical data, including the measurements of physical engine parameters as well as the measured output, emissions and pulsation. The service will be rolled out and integrated as an offering to GE's customers.

Main financing partner: Innosuisse
Project partner: General Electric (formerly Alstom Power)