Peter Lane: Abstract
Successful data mining relies on an interaction between computer
processing and
human analysis. In this talk, I consider two aspects of this interaction:
the
kind of information a data-mining system might be required to generate,
and its
interpretation. A large number of data-mining algorithms exist, with
many
focusing on classification or placing data into groups. I shall look at
some
domains where human experts require finer interpretations of data than
those
typically available. I will then consider the kinds of algorithms which
might
better emulate human reasoning, perhaps supporting more interesting
interpretations of data and more interpretable results.
|