ANN: Temporal/Causal rule generation, and more

From: Kamran Karimi (karimi_at_cs.uregina.ca)
Date: 12/18/03 

Date: Thu, 18 Dec 2003 12:39:28 -0600

Hi everybody,

 To people interested in discovering temporal/causal relations, or in
extracting classification rules from sequential data, or in generating
Prolog statements for automatic programme generation from observed
patterns, I
recommend trying TimeSleuth version 2.5. As input it accepts sequences
of
records, generated in a temporal or spatial order, and investigates
whether
there is a relation among the fields, not only in a single record, but
also
among many records. For example, it can detect if the value of a
decision
attribute is determined by the condition attributes in previous,
next, or both previous and next, records.

 With TimeSleuth time can move in either forward or backward (or both)
directions. This property is used to distinguish between causal and
acausal
relations.
 The TimeSleuth package includes executables for any platform with a
Java
virtual machine, sources in Java, online help, example datasets, and
modified C4.5 sources (plus executables for Windows). It is availabe
freely
from http://www.cs.uregina.ca/+AH4-karimi/downloads.html.

 If you have any questions, bug reports, or comments, please contact me
at
karimi@cs.uregina.ca.

Kamran Karimi

karimi@cs.uregina.ca
http://www.cs.uregina.ca/+AH4-karimi