Introduction Decision Modeling
Mis à jour :
Decision process
A decision is not an instantaneous phenomenon, but rather impact point resulting from a certain number of steps, namely, the impact point of a process.
Those steps (or highpoints) are defined by the time of a decision process which marks a break with respect to the previous period. Breaks are typically the arrival of a new information, an important event, or even the modification of the decision problem itself.
In this process, there is a progressive elaboration of options, which are fragments of decision.
Participants, stakeholders, decision-maker
An individual (group) is a participant to a decision process if, through his/her value system, he/she influences (directly or indirectly) the outcome of the decision process.
For a group of individuals to be identified as a single participant to the decision process, the value system and informational system of the members of this group should be, if not identical, sufficiently close.
The set of participants to a decision process usually evolve over time (emergence/disappearance of participants).
Decision aid / support
Definition (Decision Aid): Models to help a participant of a decision process to elaborate pieces of answers to questions he/she faces. It should increase the consistency between the evolution of the process and this participant objectives and value system.
Example
Example: A dairy collects on a daily basis the milk produced in several farms, and wants to do so in an efficicent manner.
Taken litteraly, the problem is a Traveling Salesman Problem, with the set of options being hamiltonian circuits. This model is a caricature representation, usually solved with combinatorial solutions.
Note: The model is too simplitic as:
- Farms production varies and is not known in advance.
- Transportation time is variable (weather, trafic,…), timing for cows milking is subject to âtime windowsâ.
- There exists 3 Trucks â the problem changes in its nature,
- A single criterion? risk of uncollected milk, …
The model can not be considered out of the context.
Building a decision model can contribute to modifying the formulation of the initial problem:
- Installation of a second dairy
- Policy for bying/renting trucks
- Subcontracting milk collection
Load balancing problem: Cover the load using a minimum staf
Problem data:
- $t = 1, 2, …, T$ periods
- $C_{t}$ = load for period $t$, $t = 1, 2, …, T$
- $h = 1, 2, …, H$ working hours types
- $a_{ht} = \begin{cases}1& \text{if working hours type h covers period t}\ 0 & \text{ otherwise}\end{cases}$
Decision variables: x_{h}, h = 1, .., H is the number of persons employed using working hours type h.
Optimization model:
Note: There can be additional constraints such as:
Constraints | Adjustments | Practical |
---|---|---|
Legal | Pauses | Who? |
Labor agreements | Illness + delay | When? |
Habits | Strikes | Assignment? |
Work load curve ? Usually rules like type of airplane à destination à time
Alternatives, problem stat
The alternatives are typically the possible options (hamiltonian path for traveling salesman problem for instance).
Types of alternatives:
Evolution | Solution(s) |
---|---|
Stable | Fragmentary |
Change | Holistic |
Decision problem statements:
- Choice of the best alternative
- Rank from the best to the worst (ranking can be complete or partial)
- Rank from the best to the worst (ranking can be complete or partial)
Note: More decision problem statements
“Chain” problem statements | Beyond Choice-Sort-Rank |
---|---|
Sorting + Choice | k-best |
Sorting + Ranking | Sorting with cardinality constraints |
Ranking + Sorting | Portfolio selection |
Laisser un commentaire