Handwritten Digit Recognition using Machine Learning Algorithms

Abstract

Recognition of handwritten characters and numbers is one of the practically important tasks in pattern. Numeric recognition applications include mail sorting, bank check processing, form data entry, etc. The main purpose of this paper is to demonstrate and present the work related to handwritten digit recognition, and the challenge is to be able to develop an efficient algorithm capable of recognizing handwritten digits sent users through scanners, tablets and other digital devices. In this recognition exercise, the numbers are written or written inaccurately because they differ in shape or size; for this reason, feature extraction and segmentation of handwritten numerical writing is difficult. This article presents an offline handwriting digit recognition approach based on several machine learning techniques.The main goal of this document is to provide efficient and reliable approaches to handwritten digit recognition. Several machine learning algorithms have been used for digit recognition using MNIST, most notably multilayer perceptron, and support vector machine, naive Bayes, Bayesian network, random forest, J48 and random tree.