Mining Experience with Transaction Data
Dr. Vernon Rego
Thursday, November 3, 2005
1:30PM - CSB-232
Abstract
We present experimental results on two problems involving transaction
data, where the data may contain hidden and/or misleading information.
The first study examines a certain hypothesis pertaining to strikes on
option expiry, as in aid in prediction. The results suggest that the
hypothesis is false, though it can be useful when recast under different
assumptions. The second study involves secrecy and the information
content of transaction data. In both cases we present experimental
methodology and discuss ongoing work.
Short Bio
V. Rego is a Prof. of Computer Sciences at Purdue University. His
interests are in the area of distributed computing, all aspects of
simulation, software and data/probability modeling. He was awarded a
Gordon Bell Prize and a German DFG award for distributed computing and
network protocol research.
|
