Certified Data Mining and Warehousing Professional Developing decision trees

Developing decision trees
 


A decision tree is a kind of flowchart -- a graphical representation of the process for making a decision or a series of decisions. Businesses use them to determine company policy, sometimes simply for choosing what policy is, other times as a published tool for their employees. Individuals can use decision trees to help them make difficult decisions by reducing them to a series of simpler, or less emotionally laden, choices. Regardless of the context or type of decision, the structure of a decision tree remains the same.

1 Brainstorm each of the variables in the decision you want the decision tree to help you make. Write them down on a sheet of paper, or in the margin of your main sheet.

  If you were making a decision tree for buying a car, your variables might be "price," "model," "fuel efficiency," "style" and "options."

2 Prioritize the variables you've listed and write them down in order. Depending on the kind of decision you're making, you can prioritize the variables chronologically, by order of importance, or both.

  For a simple work vehicle, you might prioritize your car decision trees as price, fuel efficiency, model, style and options. If you were buying the car as a gift for your spouse, the priorities might go style, model, options, price, fuel efficiency.

3 Draw a circle or box on 1 edge of your paper and label it to represent the most important variable in your decision tree.

  When buying a work vehicle, you might draw a circle on the left edge of your paper and label it "price."

4 Create at least 2, but preferably no more than 4, lines leading out from the first variable. Label each line to represent an option or range of options derived from that variable.

  From your "price" circle, you could draw 3 arrows labeled "under $10,000," "$10,000 to $20,000" and "over $20,000."

5 Draw circles of boxes at the end of each line, representing the next priority on your list of variables. Draw lines radiating from those circles representing the next set of options. In many cases, the specific options will be different for each box, based on the parameters chosen from your first decision.

  In our example, each box would read "fuel efficiency." Because less expensive cars often have lower gas mileage, your 2 to 4 choices from each fuel efficiency circle would represent a different range.

6  Continue adding boxes and lines to your flowchart until you've reached the end of your decision matrix.

  It's common to come up with additional variables while you're creating your decision tree. In some cases, these will apply to only 1 "branch" of your tree. In others, it will apply to all branches.

 

 For Support