KNN Classifier

A k-Nearest Neighbors supervised classifier in Python 3.11.7

Importing the package

pip install knnclassifier

from knnclassifier import KNNClassifier

This statement imports the KNNClassifier class, which will be used to create your classifier objects.

Using the package

Fitting the classifier

Set up your training and test data and fit the classifier.

1. Populate two text files with your training and test data.

1.4, 4.0, 15.0, 1.1, 1.6, 5.0
1.2, 4.0, 22.2, 2.2, 3.0, 6.0
...

Each line serves as a record and contains a list of its data points (fields), with the record's label as the final item in the list. In the above examples, the labels are 5.0 and 6.0, respectively. Ensure that fields are separated by a consistent delimiter (e.g., a comma and a space).

2. Split the data and label arrays from the files.

train_data, train_label = KNNClassifier.split_data_label_from_file("train.txt", delimiter=", ")
test_data, actual_label = KNNClassifier.split_data_label_from_file("test.txt", delimiter=", ")

The static method split_data_label_from_file returns two numpy arrays. The first contains all columns except the last (the data columns), and the second contains the last column (the label column).

You may also specify the delimiter. By default, it is ", ", but you may specify any string.

3. Initialize a KNNClassifier object and fit it to the training arrays.

classifier = KNNClassifier()
classifier.fit(train_data, train_label)

Making and checking predictions with the classifier

1. Compute and store label predictions for the test data.

test_label_predictions = classifier.predict(test_data, k=5)

The predict method returns the predicted labels of each record of the test data. Specify the k value to be used during prediction.

2. Check the accuracy of the label predictions.

accuracy: float = KNNClassifier.check_accuracy(test_label_predictions, actual_label, print_info=False)

The static method check_accuracy returns the proportion of the predicted labels that match the actual labels.

You may also specify print_info (default: False). If set to True, each label will be printed individually.