kTLNN-Python

This is a python implementation of the Two-Layer Nearest Neighbor classifier proposed by Wang et al.

Classifier description

The operation of the classifier is divided into three stages, given a query point $x$:

1. The first layer is built by the $k$ nearest neighbors of $x$. Then the second layer is formed by the $k$ nearest neighbors of each point in the first layer.
2. Considering the distribution of the query with the second layer points, the extended neighborhood is built with the most salient points in the first and second layer.
3. The final neighborhood is formed by the points in the extended neighborhood which contain $x$ among their $k_b$ nearest neighbors (backward nearest neighbor relation). The final vote to decide the $x$ class is performed on this neighborhood.

Installation

Clone this repository

git clone [email protected]:daniel-lima-lopez/kTLNN-Python.git

move to installation directory:

cd kTLNN-Python

Basic usage

Import the class and instantiate the classifier

import TLNN as tl
classifier = tl.kTLNN(k=5, kb_factor=1.4)

where:

• k is the number of neighboors considerd in the construction of layers.
• kb_factor is the $c$ factor to calculate the $k_b$ parameter, calculated as $k_b=c\cdot k$. By default this parameter has value $1.4$, as suggested by the authors.

Once instantiated the classifier, we can perform a simple test:

import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

# data preparation
dataset = pd.read_csv('Datasets/iris.csv')
X = dataset.drop('class', axis=1).values
y = dataset['class'].values

# data split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

# fit the classifier
classifier.fit(X_train, y_train)

# predictions
preds = classifier.predict(X_test)
print(f'accuracy: {accuracy_score(y_true=y_test, y_pred=preds)}')

Experiments

Experiments were performed with the datasets: balance-scale, bands and wine. On each experiment, a 10-fold cross validation was applied considering the k values 1, 3, 5, 7, 9 and 11. The accuracy on each experiment is presented in the following figures: