Transcript Decision Trees by Charles Henry

Decision Trees
Charles Henry: SE157B
Purpose
 Allow us to map out different paths for a
complex series of decisions.
 Manage decisions within our control.
 Weight outcomes based on events
beyond our control.
Nodes & Leaves
 Squares typically represent decisions.
 Circles typically represent uncontrollable
events.
 Leaves of the tree represent final
outcomes.
Problem at hand
A company wants to decide if it is financially
sound to develop new products,
focus on enhancing existing products, or if it
would be best to simply do nothing and
continue business as usual.
Contructing the Tree
 Develop a tree skeleton with decisionnodes.
 Add external chance-nodes.
 Add outcome-leaves
Tree skeleton
Establish outcomes
 Establish what the leaves will represent.
 In this case the outcome for each
scenario will be monetary gain.
Weight external events
 The market reaction is an uncontrollable
event.
 Analysts can make educated estimates
on each potential outcome.
 Label the branches with weighted
probability values
Weighted Tree
Backtrack
 Note the final monetary outcomes.
 Each outcome becomes a factor.
 The second factor is the probability value
marking the branch leading to the given
outcome.
 Multiply factors and obtain a probable
monetary value.
 Sum the net values stemming from the chancenodes to view more realistic monetary
outcomes.
Weighted Tree part 2
Leaf replacement
 Replace monetary outcome nodes with
the net value of the weighted outcomes
 Label the cost of each decision node.
 Max new outcome – costm = max realistic
profit for given decision
Reduced Tree
Results
 Costs analyzed for all reduced scenarios.
 Realistic profits analyzed for all reduced
scenarios.
 Maximum profits noted for each decision
node.
 Sound business decisions can be made.
Questions?
Sources
 http://www.time-managementguide.com/decision-tree.html
 http://www.mindtools.com/dectree.html
 http://www.spss.com/software/statistics/d
ecision-trees/index.htm?tab=1