13.txt

This complete matlab for neural network

发信人: roamingo (漫步鸥), 信区: DataMining
标  题: Soft: RFCT 3.0 for temporal/causal rule discovery
发信站: 南京大学小百合站 (Thu Jan 17 12:04:44 2002), 站内信件

[ ------ from the ML group @ [e-mail] ------ ]

Date: Wed, 16 Jan 2002 14:20:10 -0600 (CST)
From: Kamran Karimi <[e-mail]>
To: [e-mail], [e-mail]
Subject: [ML] RFCT 3.0 for temporal/causal rule discovery

Hi everybody,

 RFCT is a tool based on C4.5 and written in Java. It uses C4.5 to discover
temporal and causal rules, and has the following features:

 *) Has a graphical user interface.
 *) Handles temporal data, both in input and output.
 *) Can function in an unsupervised manne.r
 *) Outputs temporal/causal rules in a useful manner, so the user can
    have a good understanding of what influences the result.
 *) handles continous values (can discretize real-valued variables).
 *) Can output rules in Prolog, thus the rules are readily executable.

 More information about RFCT follows.

 RFCT 3.0 is a tool for the unsupervised discovery of temporal rules. It is 
based on C4.5 and extends its abilities by allowing it to understand and 
respect temporal order among the input data. RFCT's different methods of 
presentating the results help the user discover relationships among the 
variables. The user can experiment with different scenarios concerning the 
passage of time between the variable observations, and thus see if the value 
of one variable is not only  determined by other variables observed at the 
same time, but also by variables observed before. The rules generated by
C4.5 can also be converted into Prolog statements.

The package, including full source code, example files, and online help,  is 
available freely from http://www.cs.uregina.ca/~karimi/downloads.html.

RFCT is written in Java, and is runnable any where with a java interpreter
and a graphical user interface. This includes Microsoft Windows and X Window.
There is no special installation necessary. Use softrware such as WinZip
under windows and (gzip -d rfct.tgz; tar -xf rfct.tar) under UNIX to unpack
the package. In the resulting rfct-3.0 directory, just type rfct.bat to start 
the program.

You will need a properly installed C4.5 package for RFCT to function. To 
fully use RFCT's abilities, you need to patch C4.5 Release 8 and recompile
it. Microsoft Windows users can download patched and compiled C4.5 files from
the same page as the RFCT package. For others, the patch file is included in 
the  package, and you can get C4.5's sources freely from 
http://www.cse.unsw.edu.au/~quinlan . RFCT can still be used as a graphical 
user interface for C4.5 if you don't patch it.  

Once the application in running, use the online help and follow the 
"sample session" link to get started with the program.

Kamran Karimi

[e-mail]
http://www.cs.uregina.ca/~karimi

----
To UNSUBSCRIBE, send an empty message to
[e-mail]
To SUBSCRIBE, send an empty message to
[e-mail]
Archives: http://www.egroups.com/messages/machine-learning

--
Read digitally, save a tree.

※ 来源:．南京大学小百合站 bbs.nju.edu.cn．[FROM: 202.120.7.27]