Random forest is currently the most accurate of all classification techniques available. Random forest is an ensemble method that works on the principle that many weak learners can come together to make a strong prediction. In this case, the weak learner is a simple decision tree, and random forest is strong learner.
Random forest optimizes the output from many decision trees formed from samples of the same data set. In general, the higher the number of trees, the better the accuracy of the resulting random forest ensemble will be. Yet, at higher numbers, the gain in accuracy decreases. So, the analyst has to decide on the number of trees, based on the cost of implementation that he/she will face with higher numbers of trees.
The trees are combined according to a voting mechanism. The voting is based on the success criteria of each tree. The best results are gained using a weighted approach, wherein the votes are weighted, based on the accuracy of individual trees. (See Figure .) Thus, the most accurate of classification model is found.
