| Klimasauskas Group specializes in
the extreme aspects of data mining. When you hit a wall in deriving
value from your data, or you want to make sure that you are getting the
most out of it, Klimasauskas Group is the right place to come to. We specialize
in pushing data-analytic technology to extremes in performance and reliability.
We work on both commercial and classified government projects. Below are
some recent examples of projects we have worked on:
Reliable Classifier Based on a Small
Data Set.
This particular application involved predicting human behavior. There
were typically 25-100 incidents which identify both the absence, or
type of a specific behavior. Each incident included 40-90 features of
the situation. Klimasauskas group developed an approach to building
high performance statistically reliable models that maximized the use
of input data, despite the small numbers of training cases. This same
approach is applicable to a large number of situations which are input
rich, data poor, and where accurate predictions have a high value augmenting
human expertise. (See: Behavioral Applications)
Adaptive Market Profiling.
A proto-type system was developed to estimate sex, age, economic, and
educational characteristics of a TV viewer based on the programs they
watched. This used statistics from recorded viewing habits.in an on-line
adaptive system that was able to differentiate between multiple viewers
in a household, and adaptively refine the viewing profile. Results of
the profile estimates could be used to insert specific advertisements
or suggest other programs that may be of interest. Cross-marketing and
cross-selling are other areas where this approach could be effective.
(See: Marketing Applications)
Non-orthogonal Signal Decomposition.
A recursive procedure was developed to systematically identify interacting
signals in a complex composite wave, and use these to synthesize a future
prediction of selected elements of the signal. (See: Financial
Applications)
|
