Re: Multi-objective vs Stochastic!

Michael wrote:

> Hi everyone,
>
> Suppose a model is LP. However all decision variables are BINARY since
> the problem structure does impose that.
>
> It is also a multi-objective one since we have several conflicting
> objectives (criteria). That means we need to generate Pareto solutions
> by varying the objective weights, and Pareto frontiers by performing
> sensitivity and scenario analysis.
>
> If uncertainty is to be considered and Monte Carlo or Stochastic is to
> be used.
>
> Question: Is there a need to do simulation/stochastic since many (most
> or all) of the solutions/frontiers obtained from the sensitivity and
> scenario analysis would/could be the same as the ones obtained from
> simulation/stochastic?

Stochastic programming is emphatically not the same as
scenario analysis, as a good stochastic solution hedges
against a variety of possible futures. So a better way
to approach your problem is to set up a full stochastic
program with multiple (carefully selected) scenarios for
each set of objective weights.

I think you should start by reading a book on stochastic
programming, such as Birge and Louveaux or Kall and Wallace.

Hope that helps.