International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Handwritten digit recognition is a classic problem in the field of computer vision, and the MNIST dataset is one of the most common benchmarks used to test different machine-learning methods. In this study, we take a closer look at how Convolutional Neural Networks (CNNs) and Support Vector Machines (SVM) perform on this task. SVM, a traditional machine-learning technique, works by treating each image as a long list of pixel values and depends heavily on manually designed features. In contrast, CNNs can automatically learn important visual patterns such as edges, curves, and shapes directly from the raw images. To understand the strengths and weaknesses of each approach, we trained both models on the MNIST dataset and compared their performance using accuracy, precision, recall, and F1-score. Our results show that CNNs consistently outperform SVM, especially when it comes to understanding subtle variations in handwriting. This happens because CNNs are better at capturing the spatial structure of images, something traditional algorithms struggle with. Overall, the study highlights why deep learning models like CNNs have become the preferred choice for image-based tasks, offering a clear advantage over classical machine-learning methods.

Keywords: Image classification, Machine learning, Support Vector Machine (SVM), Convolutional Neural Network (CNN), Handwritten digits classification, Deep learning, Accuracy comparison, Pattern recognition, MNIST dataset, Image recognition.

Downloads: | DOI: 10.17148/IJARCCE.2025.141242