Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!hookup!news.sprintlink.net!pipex!uknet!cix.compulink.co.uk!usenet
From: jbullard@cix.compulink.co.uk ("John Bullard")
Subject: RE: Looking for Time Series Example or Articles
Message-ID: <D3wKyJ.C81@cix.compulink.co.uk>
Organization: Compulink Information eXchange
References: <17328D17CS86.RVANRAAM@bcsc02.gov.bc.ca>
Date: Sun, 12 Feb 1995 20:01:31 GMT
X-News-Software: Ameol
Lines: 5

I did some fairly simple stuff on comparing standard forecasting tools 
with NN's, and the conclusion I came to was that if the R^2 was greater 
than about 90 use standard forescasting tools, but if it is less use 
NN's. Put another way, if the random element is high or if the variables 
are non-linear, use NN's.
