International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 05 | May 2025
p-ISSN: 2395-0072
www.irjet.net
Handwritten Digit Recognition Using Logistic Regression Mr. Faizan Ahmad1, Syed Falah Jamal2, Sachin Yadav3, Gulam Mudassir Zafar4 and Faiza5 1Assistant Professor, Department of Computer Science & Engineering, Lucknow India
2Student, Department of Computer Science & Engineering, Integral University, Lucknow India 3Student, Department of Computer Science & Engineering, Integral University, Lucknow India 4Student, Department of Computer Science & Engineering, Integral University, Lucknow India
5Student, Department of Computer Science & Engineering, Integral University, Lucknow India --------------------------------------------------------------------***---------------------------------------------------------------------ABSTRACT
Handwritten digit recognition remains a complex task due to the wide range of variations in individual handwriting styles. This research aims to provide a foundation for future developments in the field by identifying and addressing the challenges associated with digit recognition. A detailed review of existing machine learning techniques was conducted to determine the most accurate and efficient methods for classification. The study used the MNIST dataset, consisting of 60,000 grayscale images of handwritten digits, each sized 28x28 pixels. These images were employed to train various models and evaluate their performance against test data. Among all the approaches analyzed, the classifier ensemble method achieved the best results, with an impressively low error rate of 0.32%. The paper offers a comparative analysis of several techniques including Convolutional Neural Networks (CNN) and Support Vector Machines (SVM), highlighting their strengths and limitations. The findings aim to guide researchers toward more accurate and reliable digit recognition systems. Keywords— CNN, MNIST, Handwritten Digit Recognition, SVM.
1. INTRODUCTION The recognition of handwritten characters has been an area of interest since the early 1980s. As defined by the Collins dictionary, a digit is a numeric symbol ranging from 0 to 9. These digits play a crucial role in various aspects of daily life. Industries such as banking, healthcare, and insurance are heavily reliant on accurate digit interpretation. For example, in banking operations—whether opening an account or processing a withdrawal— digits like account numbers and contact details are often handwritten by customers. These handwritten entries are either manually verified by staff or scanned and interpreted by machines. Similarly, in the healthcare domain, medical forms include handwritten details such as patient IDs, prescriptions, and dosage instructions—all of which must be accurately understood. Tax documents, postal codes for mail sorting, and handwriting input on digital devices are additional real-world applications where digit recognition is critical.
Fig 1: Process of Handwritten digit recognition Handwritten digit recognition has thus emerged as a significant research focus, especially within the fields of machine learning and artificial intelligence. This system aims to convert human-written numerals into machine readable form with high accuracy and reliability. Various machine learning algorithms are employed to process and classify these digits, but challenges remain due to the vast diversity in individual handwriting styles and the presence of different numeral representations across languages.
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 716