Handwriting recognitionusing machine learning

The problem of handwriting recognition has always been involved in machine learning studies. There are many stages of handwriting recognition. These are; pre-processing, segmentation, feature extraction, classification and post-processing stages. Each of these stages can be subject to a separate academic study. The subject of this study is the classification of the pre-processed and segmented data set. MNIST data set was used in this study. This data set contains 60,000 samples from 250 different people. Includes handwritten images of numbers 0 to 9. It is frequently preferred in academic studies. Many machine learning methods were run on the MNIST data set. Accuracy score was calculated for each method and results were reported. For the study, the most common machine learning methods in the field of handwriting recognition were selected. As a result of these researches and tests; the methods providing the highest efficiency from the methods compared were K-Nearest Neighbor algorithm and K-Means algorithm. Of course, the existence of many factors such as working time and data set should be taken as a criterion for those who want to work in this field. This thesis study has made a basic introduction to the field of handwriting recognition and has tested many different machine learning methods for handwriting digit recognition. It is aimed to be a guide in this study area.