A goal group is simply a collection of goals (dependency networks). Goalgroups are handy for grouping dependency networks for evaluation.
The goal group node generally organizes a group of goal nodes to either compare them simultaneously or to use the cumulative effect of the goals to build a higher level characteristic. These uses are qualitative and quantitative, respectively.
Qualitative groups are composed of discrete goals. Discrete goals are ones that are unique to themselves and not necessarily related in a symbiotic fashion to each other.
Discrete goal groups (qualitative groups) can be used to either rank or select a best goal by evaluating all the goals in the group and ranking them according to their relative trueness. In this type of group the goals may not necessarily be competing against each other, rather each goal could stand on its own as a decision.
An example of a qualitative goal group with competing goals might be an insect key, a goalgroup of insects where the purpose is to determine which one of the many insects in the group is the one of interest. Each insect would be represented by a goal that would describe the mix of characteristics that indicate the presence of the insect. Once enough information has been input, say body shape and size, coloration, etc., only one insect (goal) should be true and, if we built things well, the others should be false.
By comparison, we could build a knowledge based system to select aerobic sports. We might include as goals: biking, running, and skiing and base our decision on the available time, the weather, and equipment condition. It is conceivable that all goals in this group could be true or even all goals could be false.
Quantitative groups are composed of goals (dependency networks) that represent positions along a continuum of possible values. (The position held by a goal is determined from the weight associated with the goal.) Goals that are positions on a continuum represent specific positions relative to each other goal in the goalgroup and are used to return a continuous value.
As an example, the group “Dangerous” might include the goals “Danger.low”, “Danger.moderate”, and “Danger.high”. The group would be used to select the best point along the danger continuum from low to high.
Quantitative goal groups are used to determine the point on the continuum that best fits the given data using the fuzzy logic technique of finding the centroid of the goalgroup. The centroid is calculated using the weights associated with the goals as their center points along the x-axis and the trueness levels of the goals as their heights. All the goals in the group contribute to the centroid calculation.
The qualitative goalgroup's value is index number of the most true goal in the group:
value = i, where i is index of max(value1, value2, … , valuen)
The quantitative goalgroup's value is the centroid of the values of the goals (goals as points on a continuum). The centroid is calculated by dividing the sum of moments by the sum of magnitudes. The moments are the magnitudes multiplied by the weights (weights represent the position of the goal along the continuum).
value = [ (value1 + 1)weight1 + (value2 + 1)weight2 + … + (valuen + 1)weightn ] / [ (value1 + 1) + (value2 + 1) + … + (valuen + 1) ]
An example of a quantitative goalgroup could be a risk rating system where we want to determine the level of risk associated with a given scenario. The group could have goals for definable levels of risk: Risk.vhigh, Risk.high, Risk.mod, Risk.low, Risk.vlow. The goalgroup would evaluate each goal and then combine the results to give a single value that represents the center of mass of the area under the curve defined by the individual goals.
In the example below a moderate risk (Risk.mod) has been determined.
Risk Evaluation | ||||
---|---|---|---|---|
goal | Weight | Value | Magnitude | Moment |
Risk.vLow | 1.00 | 100% False | 0.000 | 0.000 |
Risk.low | 2.00 | 75% True | 1.750 | 3.500 |
Risk.mod | 3.00 | 100% True | 2.000 | 6.000 |
Risk.high | 4.00 | Undetermined | 1.000 | 4.000 |
Risk.vhigh | 5.00 | 45% False | 0.550 | 2.750 |
Sums | 5.300 | 16.250 | ||
Centroid | 3.066 |