KNOWLEDGE MANAGEMENT

AND DISCOVERY LAB

WHAT IS DIFFERENT FROM CONVENTIONAL MODEL LEARNING? When a model is inferred, it reflects the characteristics of the population it was built upon. But the population - be it user preferences in the web, be patients arriving in a hospital clinic, be it energy consumption of a town, be it the work accomplished by a machine in an industrial floor - this population is changing. It is changing all the time, due to external phenomena and due to internal properties.

• Internal properties: people become older and machines as well. For humans, this affects their preferences for products, the likelihood that they can get some ailment (some ailments are characteristic of old age, such as changes in the eyesight). For machines, this corresponds to fatigue of the materials.
• External phenomena: user preferences are affected by new products that enter the market; the influx of patients at a cliniic can increase when the weather becomes suddenly very cold; an outage in an electricity network (eg because of a machine that went out of order) will lead to a peak in demand when the network becomes availale again.

The consequence is that something must be done: predict these changes and anticipate them. This leads to the tasks of model adaption and of prediction of the future. In DM II, we discuss these two tasks.

 Veranstaltungsbegin ist 9:30 Uhr.

Scientific papers (to be announced at the course)

Data Mining (recommended)

WPF Master DKE

WPF Master Inf

WPF Master WIF

WPF Master CV

WPF Master IngInf

WPF Master Statistik