2.3. Unsupervised Learning — Learning Without Answers

The computer is given data with no labels or answers. It must figure out patterns on its own. Imagine giving a kid a basket of mixed fruits — apples, bananas, and oranges — without telling what’s what. If the kid groups them by shape or color, that’s unsupervised learning.

Unsupervised learning is a type of machine learning where the algorithm is given data that has not been labeled or categorized beforehand. The goal is for the algorithm to identify patterns, similarities, or differences in the data autonomously, without any guidance on what those patterns might be. For instance, if you present a set of mixed fruits to a child, they must use their observation skills to separate the fruits based on their characteristics such as shape or color, rather than being told which fruit is which.

Think of unsupervised learning like a teacher who gives a class a puzzle without showing them the completed picture. The students must work together to figure out how to fit the pieces together based solely on the shapes, colors, and patterns of the individual pieces.

● Group similar things (Clustering)
● Find hidden structure
● Detect unusual items (Anomaly Detection)

Unsupervised learning has various applications, mainly centered around identifying and interpreting patterns in data. This can include clustering, where the algorithm groups similar items together (like different types of fruits or customers with similar spending habits). It can also unveil hidden structures within a dataset, such as identifying groups of students with similar learning styles. Additionally, it is adept at detecting anomalies, or outliers, which are unexpected items that differ significantly from the rest, like fraudulent transactions among regular purchases.

Imagine you are a librarian who has just received a large number of books in various languages and genres. Without categorizing them initially, you might notice that certain books are similar based on their cover colors, authors, or themes. By organizing them into groups, you can discover patterns such as 'sci-fi novels' or 'children's books', making it easier to find what you're looking for later.

from sklearn.cluster import KMeans
import numpy as np

Each pair = spending and visits

data = np.array([
[1, 2], [1, 4], [1, 0],
[10, 2], [10, 4], [10, 0]
])
model = KMeans(n_clusters=2)
model.fit(data)
print("Cluster labels:", model.labels_)
print("Centers:", model.cluster_centers_)

In this example, we use a KMeans clustering algorithm to classify customers based on their spending habits and visit frequency. The data represents two features: how much each customer spends and how often they visit. By fitting the model to this data, it groups the customers into clusters – in this case, potentially separating low-spending customers from high-spending ones. The model identifies core 'centers' for each cluster, demonstrating the average behaviors of customers within those groups.

Think about a store owner who wants to understand their customers better. By grouping them based on the amounts they spend during visits, the owner can identify different customer segments, like budget shoppers and luxury buyers, allowing them to tailor marketing strategies to each group's preferences.

--