Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!agate!ames!waikato!canterbury.ac.nz!otago.ac.nz!is30224
From: is30224@otago.ac.nz
Subject: Algorithms, etc for effort estimation
Message-ID: <1994Oct7.194244.1@otago.ac.nz>
Lines: 17
Sender: usenet@news.otago.ac.nz (News stuff)
Nntp-Posting-Host: velos.otago.ac.nz
Organization: University of Otago, Dunedin, New Zealand
Date: Fri, 7 Oct 1994 06:42:44 GMT

Hi Everyone!

	I am currently carrying out some research using neural networks to
estimate software engineering effort.  The inputs are specification-based
complexity indicators from automated development environments (CASE and/or
4GL's).  For example total number of read transactions, total number of
distinct screens displayed, total number of entities, etc.  The outputs are
effort measures such as analysis-design time, total time, etc.
	I have been using backpropagation for my work, and the results have
been encouraging.  However I am interested in any suggestions as to different
algorithms to use (I am currently looking at cascade-correlation).  Any other
suggestions would also be welcomed.  The fundamental problem is to estimate the
expected effort given the complexity measures.
	If you reply to me via email then I will summarise to the group.
	
Thanks in advance,
Andrew Gray
