MNIST Digit Recognition Project

This project performs digit recognition on the MNIST dataset using Decision Tree and Support Vector Machine (SVM) models.

Features

• Dataset: 42,000 samples of 28x28 pixel grayscale images of handwritten digits (0-9). Each sample is represented by 784 pixel intensity values.
• Preprocessing: Min-Max normalization and a 75%-25% train-test split.
• Models:
  • Decision Tree Classifier (max depth=15)
  • Support Vector Machine (SVM) with RBF kernel
• Evaluation: Accuracy, F1 score, and confusion matrix for model performance.
• Visualizations:
  • Class distribution using bar and pie charts
  • Sample images from the dataset
  • Confusion matrices for model evaluation
  • Decision Tree structure visualization

Requirements

• numpy
• pandas
• matplotlib
• seaborn
• plotly
• scikit-learn
• graphviz
• pydotplus

Instructions

1. Place the mnist.csv dataset in the project directory.
2. Optionally, test the trained models with a new dataset (test.csv).

---