k-nearest neighbor algorithm for supervised learning in Python

So far we have been going with the prerequisites to Machine Learning. We have covered a few of the important concepts that are required to solve Machine Learning problems. If you have missed those posts, I recommend you to go through Machine Learning Concepts.

Here we will be focusing on Supervised Machine Learning. If you are not sure about various techniques to solve machine learning problems. 

Let's get started.

Supervised Learning

is the learning where the value or result that we want to predict is within the training data (labeled data). The value which is in the data that we want to study is known as Target or Dependent Variable or Response Variable.

All the other columns in the dataset are known as the Feature/Predictor Variable or Independent Variable.

Supervised Learning is classified into two categories:

1. Clarification: Here our target variable consists of the categories.

2. Regression: Here our target variable is continuous and we usually try to find out the line of the curve.

As we have understood to carry out supervised learning we need labeled data. How we can get labeled data? There are various ways to get labeled data:

1. Historical labeled Data

2. Experiment to get data: We can perform experiments to generate labeled data like A/B Testing

3. Crowd-sourcing

Now it's time to understand algorithms that can be used to solve supervised machine learning problems. In this post, we will be using the popular scikit-learn package.

NOTE: There are few other packages as well like TensorFlow, Keras etc to perform supervised learning.

k-nearest neighbor algorithm

This algorithm is used to solve the classification model problems. K-nearest neighbor or K-NN algorithm creates an imaginary boundary to classify the data. When new data points come in, the algorithm will try to predict that to the nearest of the boundary line.

Therefore, a larger kvalue means smother curves of separation resulting in less complex models. Whereas, a smaller k value tends to overfit the data and result in complex models.

NOTE: It's very important to have the right k-value when analyzing the dataset to avoid overfitting and underfitting of the dataset.

Using the k-nearest neighbor algorithm we fit the historical data (or train the model) and predict the future.

Example of the k-nearest neighbor algorithm

Consider the example shown below:

# Import necessary modules
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.datasets import load_iris

# Loading data
irisData = load_iris()

# Create feature and target arrays
X = irisData.data
y = irisData.target

# Split into training and test set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state=42)

knn = KNeighborsClassifier(n_neighbors=7)

knn.fit(X_train,y_train)

# Predict on dataset which model has not seen before
print(knn.predict(X_test))

In the example shown above following steps are performed:

1. The k-nearest neighbor algorithm is imported from the scikit-learn package.
2. Create feature and target variables.
3. Split data into training and test data.
4. Generate a k-NN model using the neighbor's value.
5. Train or fit the data into the model.
6. Predict the future.

We have seen how we can use the K-NN algorithm to solve the supervised machine-learning problem. But how to measure the accuracy of the model?

Consider an example shown below where we predicted the performance of the above model:

# Import necessary modules
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.datasets import load_iris

# Loading data
irisData = load_iris()

# Create feature and target arrays
X = irisData.data
y = irisData.target

# Split into training and test set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state=42)

knn = KNeighborsClassifier(n_neighbors=7)

knn.fit(X_train,y_train)

# Calculate the accuracy of the model
print(knn.score(X_test, y_test))

Model Accuracy

So far so good. But how to decide the right k-value for the dataset? We need to be familiar with data to get the range of the expected k-value but to get the exact k-value we need to test the model for every expected k-value. Refer to the example shown below:

# Import necessary modules
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.datasets import load_iris
import numpy as np
import matplotlib.pyplot as plt

irisData = load_iris()

# Create feature and target arrays
X = irisData.data
y = irisData.target

# Split into training and test set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state=42)

neighbors = np.arange(1, 9)
train_accuracy = np.empty(len(neighbors))
test_accuracy = np.empty(len(neighbors))

# Loop over K values
for i, k in enumerate(neighbors):
    knn = KNeighborsClassifier(n_neighbors=k)
    knn.fit(X_train,y_train)
    
    # Compute traning and test data accuracy
    train_accuracy[i] = knn.score(X_train,y_train)
    test_accuracy[i] = knn.score(X_test,y_test)

# Generate plot
plt.plot(neighbors, test_accuracy, label = 'Testing dataset Accuracy')
plt.plot(neighbors, train_accuracy, label = 'Training dataset Accuracy')
plt.legend()
plt.xlabel('n_neighbors')
plt.ylabel('Accuracy')
plt.show()

OUTPUT

Here in the example shown above, we are creating a plot to see the k-value for which we have high accuracy.

NOTE: This is a technique which is not used industry-wide to choose the correct value of n_neighbors. Instead, we do hyperparameter tuning to choose the value that gives the best performance. We will be covering this in future posts.

Summary

In this post, we have understood what supervised learning is and what are its categories. After having a basic understanding of Supervised learning we explored the k-nearest neighbor algorithm which is used to solve supervised machine learning problems. We also explored measuring the accuracy of the model.