Machine Learning

Digit Recognizer

A pipeline that classifies handwritten digit images (MNIST) using PCA for dimensionality reduction and supervised classifiers (Decision Tree, SVM).

Feb 2024 Eng. Ameen Qahtan

Project Overview

The project loads the MNIST training set with pandas, applies PCA to preserve 95% of variance, then fits both a Decision Tree and an SVM (with hyperparameter tuning via GridSearchCV). Final evaluation reports an overall accuracy of 85% along with per-class precision, recall, and F1-scores .