Impact of Machine Learning and Deep Learning Models on Handwritten Digits and Letters Recognition (HDaLR)
Year: 2024
Type: article
Abstract: In numerous practical applications, such as data form entry, postal code sorting, and bank check account processing, handwritten digit recognition is one of the crucial and difficult tasks. Because each person writes in a distinct way with varying sizes, widths, and slopes, it can be challenging to recognise digits. Various artificial neural network-based models have been used in the past for pattern matching. While conducting the experiment, significant differences in the use of fonts by various authors were observed using the MNIST (Modified National Institute of Standards and Technology database) dataset as a benchmark. In this study, we evaluated machine learning algorithms on the MNIST dataset, including Naive Bayes, K-Nearest Neighbor, Support Vector Machine, Decision Tree, Random Forest, Artificial Neural Network, Convolution Neural Network, and Long Short-Term Memory. The purpose of this research is to evaluate and contrast the effectiveness of deep learning and machine learning models over handwritten letters and digits datasets. It was noted that CNN has outperformed, and the accuracy obtained is 99.9% over the MNIST dataset and 88% over the EMNIST dataset. Every identification approach faces the crucial challenge of extracting key features, and deep learning has been used to solve this problem with results that have been evaluated.
Institution: Savitribai Phule Pune University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4391034067