Handwriting digit recognition using Decision Tree Classifiers

Dublin Core


Handwriting digit recognition using Decision Tree Classifiers


Demir Korać, Samed Jukić, Mujo Hadžimehanović


The usage of handwritten character recognition has been useful for usage from large to
common consumer usage. The transitional period of the handwritten to the digital age can be
largely improved by focusing on perfecting handwritten character recognition.
This paper and work aims to focus on handwritten digit recognition using the decision tree
classifier machine learning method, implemented, trained and tested on the data set gathered from
the Modified National Institute of Standards and Technology dataset.
The data to be recognized is inputted from a pre-existing reliable set, used both for training and
testing, in order to give a fair result. The system is run through a Python script and the data set is
stored in CSV format, preprocessed and ready for further usage.
Taking into consideration the size of the dataset (42000 rows of data), the system’s overall
performance is satisfactory with an accuracy of 85% and outputs the results in an understandable


character, decision, handwritten, recognition


ISSN 2637-2835


International Burch University, Sarajevo, Bosnia and Herzegovina


Journal of Natural Sciences and Engineering


January, 2020

Document Viewer