Digit-Interpreter

Overview

Classification algorithm for interpreting digits using Python and MNIST dataset based on k Nearest Neighbours machine learning algorithm implemented from scratch.

1. Requirements

• Python 3.x

2. Build and run setup

1. Download MNIST dataset e.g. from https://www.kaggle.com/oddrationale/mnist-in-csv and extract it to project main folder
2. Run project. When process end inside project folder will appear two files:

• classification.csv - contain sets of real vaule digits and matching them digits predicted by algorithm for k=1..k neighbours
• errors.csv - contain total amount of bad predictions for k=1..k neighbours Also process will draw chart showing statistics total number of good predictions to bad predictions depending on number of neighbours.