Re: Can OR be applied here.

"Shashank Khanvilkar" <[EMAIL PROTECTED]> wrote in
news:[EMAIL PROTECTED]: 

> Hi,
> Will appreciate some help here... (note: newbie in OR).
> 
> I have a practical problem at hand which is described as followss
> 
> 1. I have multiple implementations for a software system (say a simple
> Data-base system), provided by different vendors (say mysql,
> postgresql etc.).
> I have to select only a single implementation (or narrow down my
> options) from this set. To compare them I have found out some
> attributes that I want to compare and assigned a score to each
> implementation for that attribute. 
> 
> for e.g. I have the following table
> ----------------------------------------------------------
> Attributes     |        MySql     |     PostGreSQL
> ----------------------------------------------------------
> X1                  |        8               |      7
> X2                  | 5                      |      6
> X3                  |      7                 |      2
> etc..
> 
> Since there are many implementations and many attributes, I am not
> able to get the optimum solution form the data obtained.  Is there any
> method to do this.?
> Any help appreciated.
> 
> 

There are quite a few approaches to a problem of this type:

* Assign weights to each attribute, obtain a weighted composite score for 
each alternative, and pick a winner.  Picking the weights is an adventure 
unto itself.

* Find the Pareto optimal set of alternatives (those alternatives not 
dominated by other alternatives, nor by "averages" of other 
alternatives), weed out the dominated solutions, and then pick one 
whimsically.

* Data Envelopment Analysis could perhaps be used to select "efficient" 
choices.  DEA works with output/input ratios -- you might define all 
usability attributes and capacities as outputs, and things like 
acquisition cost, staff costs, hardware costs etc. as inputs.

* The Analytic Hierarchy Process steers you first into structuring your 
criteria, then into doing pairwise comparisons of the importance of 
criteria, then into pairwise comparisons of the alternative systems on 
each criterion, and ultimately (and somewhat magically) ends up with a 
composite score for each alternative.

* Multiattribute Utility Theory is another way to form a composite 
measure of the value of each alternative.  It allows you to incorporate 
nonlinear utilities for measurable performance characteristics. For 
instance, you might be really geeked about getting 10,000 transactions 
per second out of your database, but not as geeked about pushing that to 
20,000 TPS.  (Caveat:  If I remember correctly, there's a subtle 
distinction between utility functions and value functions, so you might 
actually need a multiattribute value function.  I'll let someone else 
who's more familiar with those things pick that particular nit.  I can 
barely spell multiattribute.)

-- Paul

*************************************************************************
Paul A. Rubin                                  Phone:    (517) 432-3509
Department of Management                       Fax:      (517) 432-1111
The Eli Broad Graduate School of Management    E-mail:   [EMAIL PROTECTED]
Michigan State University                      http://www.msu.edu/~rubin/
East Lansing, MI  48824-1122  (USA)
*************************************************************************
Mathematicians are like Frenchmen:  whenever you say something to them,
they translate it into their own language, and at once it is something
entirely different.                                    J. W. v. GOETHE