ZAbbas
0
Hi I'm a beginner to random forest algorithm in Machine Learning.
According to what I have read in theory, it works on majority vote concept for classification problems. But can it be possible that number of "Yes" is same as number of "No"?
What would be done in that case?
Can it be possible that number of "Yes" is same as number of "No"?
Decision trees do not "coordinate" their predictions amongst themselves, so any ratio of "Yes" and "No" predictions is possible, including a tie (ie. a 50/50 ratio).
What would be done in that case?
You break a tie according to a predefined tie breaking algorithm. For example, you can stipulate that the "Yes" class always wins over the "No" class in such a scenario.
It's advisable to use an odd number of decision trees in a random forest in order to eliminate the possibility of ties. For example, if your random forest has 101 member decision trees, then there can't be a 50/50 outcome - one class will always have one extra vote.