This topic covers all elements of Deterioration Modelling in JunoViewer Web
Selecting Optimal Candidate Strategies: BCA Models
|
Fritz Jooste Administrator Posts: 81
5/20/2020
|
Fritz JoosteAdministrator Posts: 81
As explained in this post, Strategy Generation is one of the first steps in the process of running a Benefit Cost Analysis. In most cases, the strategy generation algorithm will generate many strategies that are feasible but not economically optimal. JunoViewer uses the concept of a Pareto Efficiency to select a subset of optimal candidate projects on each modelling segment. This post deals with the concept of Pareto Efficiency in the context of strategy selection.
Note that: in this discussion, we are always concerned with strategies generated on a single modelling segment. Thus strategies are mutually exclusive (we can pick only one). More about this at the end of this post.
The concept of Pareto Efficiency is well known in economic and engineering disciplines. For example, consider a portfolio of candidate investment projects as shown in the figure below. All points on the graph are candidate projects that could be selected. However, in a selection where cost and benefit are important, project A is clearly superior to project B because for the same cost it yields a higher benefit. Similarly, project A is superior to project C because it yields the same benefit for a lower cost.
Projects such as project A shown above are said to be Pareto Optimal. Project A is said to "dominate" projects B and C. Projects such as project A which are not dominated by any other projects, are said to lie on the Pareto Front (also referred to as the Efficiency Font in some economic analyses). Thus all the projects shown in red in the figure above are on the Pareto Front.
So which project on the Pareto Front should be selected? This appears to be an issue of ongoing debate and is clearly dependent on some subjective or objective goals. Consider for example the set of candidate projects shown below. If I have a unconstrained funding and my objective is to maximize benefits, I will choose project Z. However, if my funding is constrained, I will rather choose project X because it yields the best "bang for buck", as indicated by the dotted blue line.
When JunoViewer selects candidate strategies for optimization, it initially will select all projects on the Pareto Front. These projects are all put forward into the Global Optimization Set. A set of heuristics is then used to select a sub-set of projects on the Pareto Front based on the modelling type. These heuristics can be roughly stated as follows:
- If you are using the NPV (or "optimal IBCR") model, the selected subset as well as the global optimizing routine will favour projects that maximize benefit (i.e. leaning towards project Z as illustrated above).
- If you are using the MaxBCR method, then the selected subset and the global optimizer will favour projects with a higher benefit cost ratio (i.e. leaning towards project X as illustrated above).
As said earlier, all of the above discussion deals with candidate strategies generated on a single modelling segment. Eventually, the model can choose only one of the strategies on the Pareto Front, but it makes available a larger set of candidates for selection in the optimization stage. Which strategy is eventually chosen on a specific segment depends on many factors but mainly on the available budget in different years and the relative benefit and cost of competing projects on other segments.
edited by Fritz on 9/22/2022
|
|
|
0
link
|
