An Overview of Results Analytics

Results Analytics is a a trade-off analytics and decision-support tool. Results Analytics allows you to visualize and compare large amounts of data, such as data from simulations driven by a design of experiments. You can then start to narrow down your data set be defining objectives and perform a trade-off analysis to arrive at a recommendation for the best option.

The difficulty of selecting the best alternative from your data set is always increased by the presence of more than one criterion. Results Analytics uses a Multi-Criteria Decision Making (MCDM) process to rank data points. Results Analytics cannot arrive at a single optimal solution to an MCDM problem without incorporating preference information. The concept of a single optimal solution is replaced by the set of non-dominated solutions where it is impossible to move away from one solution to another without sacrificing in at least one criterion. In general, the set of non-dominated solutions is too large to present to the decision maker and expect them to arrive at a final choice. Results Analytics helps you focus on your preferred solutions by applying preferences (priority and weight) and allowing you to observe how the preferences affect the ranking of the alternatives (quick "tradeoff" studies).

You use Results Analytics to define the requirements and priorities from amongst your variables; to conduct a trade-off analysis; to select the best option, alternative, or choice; or even to predict a better option. For example, you can use Results Analytics to select one or more options from a large number of alternatives. Each alternative is composed of a set of parameters, such as the fuel consumption, weight, average cost per mile, cargo capacity, price, and top speed of a set of, say, 100 car designs. You can assign criteria to these parameters, and Results Analytics uses the criteria to score each alternative and to rank them relative to each other. To score the alternatives, Results Analytics uses a ranking engine based on a multiple-criteria decision analysis (MCDA) theory.

You can specify the following criteria:

Thresholds
A threshold specifies that the value of the parameter must be greater or less than a specified number, or the value must lie between a specified range. For example, the cargo capacity of the car must be greater than 30, and the average cost per mile must be less than 0.75.
Objectives
An objective specifies that the value of the parameter must be minimized or maximized, For example, the weight and the cost of the car must be minimized; the range and the top speed must be maximized. An objective can also be a target value. For example, a target for the cargo capacity could be 20 ft³.
Priorities
A priority specifies the relative importance of a parameter and its associated objective, where a priority of 1 is higher than a priority of 5 (the default value). You assign a priority from a set of predefined categories based on the preferences agreed amongst the collaborators. Results Analytics uses the priority value to initialize the weight, which is then used for scoring and ranking the alternatives. When you assign a priority, you do not have prior knowledge of the effect of the priority on the ranking. For example, you might decide that maximizing the top speed of the vehicle is more important than minimizing the weight and assign priorities of 1 and 3, respectively.
Weights
A weight is similar to a priority in that it specifies the relative importance of a parameter, where a weight of 10 is higher than a weight of 1. However, unlike a priority, to which you assign a category, you adjust a weight to the desired value along a sliding scale. In addition, you adjust a weight after applying the initial scoring and ranking. You (and your collaborators) adjust the weights to gain a better understanding of the effects of tradeoffs and how the parameters combine to influence the scoring and ranking. For example, relaxing the weight of one parameter, say the cost of the vehicle, may significantly increase the influence of a second parameter, say the fuel consumption.

In most trade-off studies, there will be conflicting relationships between parameters. The optimum alternative is difficult to define, and the objective becomes a set of compromises or trade-offs—the Pareto Optimal set—from which you will select the best option. For example, increasing the horsepower of the engine will increase the top speed, but it will also increase the cost (which we are trying to minimize). Results Analytics allows you to explore the design space and the relationships between parameters. Results Analytics also provides tools that allow you to experiment with the decision criteria and to view the effect on the desirability of each alternative. For example, what is the effect of making the cost of the vehicle more important than the top speed?