A decision tree is a logical structure of a problem in terms fo the sequences of decisions and outcomes of chances events. It visualizes possible paths for each outcome.
Suppose we’re producing memes, we can choose to either outsource the production or produce it in-house. If outsource, then the expected profit/unit is $10. Else then we have two cases: 40% low demand, which yields $7 profit/unit or 60% high demand for $15 profit/unit.
We draw the decision tree, and it looks like this:
graph LR A>A]-->|Produce|B((B)) A-->|Outsource|C[$10] B-->|60%: High demand|D[$15] B-->|40%: Low demand|E[$7]
Where [A] is a decision node and (B) is a chance node. The square nodes are outcome nodes.
This is a back track step where we compute the expected value of the outcomes at each chance node. Then select the best option (best expected value fo outcomes) at each decision node.
Repeat this process until we reach the root node.
Whiel rolling back, when we reach a decision node, we should eliminate the decisions that has a lower expected value.
At the end, we have conditional decisions optimized based on the outcome of the chance nodes.