Comparative Analysis of Handwritten Urdu Digits Recognition through Machine Learning and Deep Learning Methods

This study presents a comparative analysis of Urdu digits recognition using Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Decision Trees, KNearest Neighbors (KNN), and the VGG architecture. Evaluating these models on a dataset of handwritten Urdu digits, we assess their performance based on accuracy, precision, recall, F1-score, and computational efficiency. The deep learning models, CNN and VGG, leverage convolutional layers to extract spatial features, while RNN explores sequential data applicability. Traditional machine learning methods, Decision Tree and KNN, serve as baselines. Experimental results demonstrate that CNN and VGG significantly outperform the traditional methods in accuracy and robustness, although with higher computational costs and training times. This study underscores the trade-offs between model complexity and performance, providing valuable insights for practitioners and researchers in handwritten digit recognition.