Improvement of Random Forest Classifier through Localization of Persian Handwritten OCR

The random forest (RF) classifier is an ensemble classifier derived from decision tree idea. However the parallel operations of several classifiers along with use of randomness in sample and feature selection has made the random forest a very strong classifier with accuracy rates comparable to most of currently used classifiers. Although, the use of random forest on handwritten digits has been considered before, in this paper RF is applied in recognizing Persian handwritten characters. Trying to improve the recognition rate, we suggest converting the structure of decision trees from a binary tree to a multi branch tree. The improvement gained this way proves the applicability of the idea.