We are using an random forest classifier to classify a set of data using training data set. The random forest requires calibration of: nos. of random trees and and nos. of the features. Identifying the best feature through random selection of features forms the basis of the best split of the node while growing the trees. But, how we can fuse the predictions?
Could you please explain what you mean by fuse the predictions? The results of the classification? I will try to give a general overview of the classification algorithm:
Random Forest (RF) uses multiple decisions trees to take the decision. These trees are built selecting, for each node, the feature that best split the classes of the observations After this training phase, in the prediction phase, the probability that an observation belongs to a class is the number of trees that classified this observations in that class divided by the total number of trees.
Best Regards,
If you wonder how final classification is obtained from all fitted trees, it's usually via "majority vote". Example: 5 trees, binary classification problem (0/1). 3 trees classify observation as "0", 2 classify observation as "1". Overall the observation will be classified as "0".
See point 3 here: http://www.bios.unc.edu/~dzeng/BIOS740/randomforest.pdf
You may refer to
cs-people.bu.edu/ajjendj/documents/AjjenJoshiR2.pdf
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs