Abstract
Definitions of support and confidence in traditional association rule
mining problems are related to frequency counts of itemsets. In some
situations, however, we may want to introduce a time factor into the
definitions of support and confidence. For example, if a securities
investor is informed that "stock ABC drops in 95% of the time when stock
XYZ rises", then the investor should remain alert if the price of XYZ
begins to go up.
We propose a new research problem of mining frequency counts of itemsets
from "sensor set data", which refers to the data continuously generated by
a set of sensors. In the example above, the price of a stock can be
regarded as the value of a sensor, which monitors the status of a
particular stock and is updated from time to time.
In the talk, we will give a definition of the sensor set data mining
problem and cover some initial thoughts on how the problem can be solved
efficiently.