6. Unsupervised Learning (Clustering)

Unsupervised learning

• Finds pattern in data (without a specific prediction task in mind)
• Uncovering hidden pattern from unlabelled data
• E.g. clustering customers by their purchase patterns
• Unlike supervised learning, supervised learning finds patterns for a prediction task

Arrays, features & samples, Dimension

• 2D NumPy array # Array store values of same data type
• Columns are measurements (the features)
• Rows represent the samples
• Dimension refers to the number of features

K-means clustering

• Finds clusters of samples & number of clusters need to be specified
• New samples can be assigned to existing clusters
• K-means remembers the mean of each cluster (the “centroids”)
• Finds the nearest centroid to each new sample
• Implemented in sklearn (“scikit-learn”)
• Example:

from sklearn.cluster import KMeans # Import KMeans

model = KMeans(n_clusters=3) # Create a KMeans instance with 3 clusters

model.fit(samples) # Pass the array of a sample to the predict method of the k-mean model

labels = model.predict(samples)

print(labels)

Scatter plots

• Scatter plot of sepal length vs petal length
• Colour points by cluster labels
• Example:

import matplotlib.pyplot as plt # Import pyplot

xs = samples[:, 0] # Assign the columns of samples

ys = samples[:, 1] # Assign the columns of samples

plt.scatter(xs, ys, c=labels,alpha=0.5) # Make a scatter plot of xs and ys, using labels to define the colours

centroids = model.cluster_centers_ # Assign the cluster centers

centroids_x = centroids[:,0] # Assign the columns of centroids

centroids_y = centroids[:,1] # Assign the columns of centroids

plt.scatter(centroids_x, centroids_y, marker='D', s=50)

plt.show()