Implementation of k – Nearest Neighbor classifier from scratch to classify data from the famous IRIS dataset of scikit-learn.
Some lines of the code have been commented deliberately. You can uncomment them to compare the accuracy of the model with default kNN implementation provided by scikit-learn.
Among all well-known ML classifiers, kNN is found to be performing best on this perticular dataset.
For 80:20 train-test split, kNN model has accuracy of 93.3%.