K-Means Clustering Using Boston Housing Dataset

K-Means Clustering Using Boston Housing Dataset

Introduction

K-Means Clustering is an unsupervised machine-learning algorithm that divides data into a predefined number of groups called clusters.

In this example, the Boston Housing Dataset is used to demonstrate K-Means clustering. The 13 numerical housing features are standardized before applying the algorithm.

Important: MEDV is not used for clustering because it is the target variable in the traditional Boston Housing dataset.

1

Load Dataset

Load the Boston Housing CSV file using Pandas.

2

Select Features

Select the 13 housing-related numerical features.

3

Standardize Data

StandardScaler converts the features to a comparable scale before calculating distances.

4

Find Suitable K

Use the Elbow Method to examine different numbers of clusters.

5

Apply K-Means

Create the K-Means model and fit it to the standardized data.

6

Assign Clusters

Each housing record receives a cluster label.

7

Analyze Clusters

Calculate the average feature values for every cluster.

8

Visualize

PCA is used to display the high-dimensional clusters in two dimensions.

Python Program

import pandas as pd import matplotlib.pyplot as plt from sklearn.preprocessing import StandardScaler from sklearn.cluster import KMeans from sklearn.decomposition import PCA # ----------------------------------------- # 1. LOAD DATASET # ----------------------------------------- df = pd.read_csv("boston-housing.csv") print("Dataset Shape:") print(df.shape) # ----------------------------------------- # 2. SELECT FEATURES # ----------------------------------------- features = [ "CRIM", "ZN", "INDUS", "CHAS", "NOX", "RM", "AGE", "DIS", "RAD", "TAX", "PTRATIO", "B", "LSTAT" ] X = df[features] # ----------------------------------------- # 3. STANDARDIZE FEATURES # ----------------------------------------- scaler = StandardScaler() X_scaled = scaler.fit_transform(X) # ----------------------------------------- # 4. ELBOW METHOD # ----------------------------------------- inertia = [] for k in range(2, 11): model = KMeans( n_clusters=k, random_state=42, n_init=10 ) model.fit(X_scaled) inertia.append(model.inertia_) plt.figure(figsize=(8, 5)) plt.plot( range(2, 11), inertia, marker="o" ) plt.xlabel("Number of Clusters (K)") plt.ylabel("Inertia") plt.title("Elbow Method") plt.grid(True) plt.show() # ----------------------------------------- # 5. CREATE K-MEANS MODEL # ----------------------------------------- kmeans = KMeans( n_clusters=3, random_state=42, n_init=10 ) # ----------------------------------------- # 6. CREATE CLUSTERS # ----------------------------------------- df["Cluster"] = kmeans.fit_predict(X_scaled) # ----------------------------------------- # 7. CLUSTER COUNT # ----------------------------------------- print("\nCluster Counts:") print( df["Cluster"] .value_counts() .sort_index() ) # ----------------------------------------- # 8. CLUSTER SUMMARY # ----------------------------------------- summary = df.groupby("Cluster")[features].mean() print("\nCluster Summary:") print(summary) # ----------------------------------------- # 9. PCA # ----------------------------------------- pca = PCA(n_components=2) X_pca = pca.fit_transform(X_scaled) # ----------------------------------------- # 10. VISUALIZE CLUSTERS # ----------------------------------------- plt.figure(figsize=(9, 6)) plt.scatter( X_pca[:, 0], X_pca[:, 1], c=df["Cluster"], cmap="viridis", s=45, alpha=0.8 ) plt.xlabel("Principal Component 1") plt.ylabel("Principal Component 2") plt.title( "Boston Housing K-Means Clusters" ) plt.colorbar( label="Cluster" ) plt.grid(True) plt.show() # ----------------------------------------- # 11. SAVE RESULT # ----------------------------------------- df.to_csv( "boston_housing_kmeans_clusters.csv", index=False ) print("\nClustered dataset saved successfully.")

Sample Output

Dataset Shape: (506, 14) Cluster Counts: Cluster 0 151 1 182 2 173 Name: count, dtype: int64 Cluster Summary: CRIM ZN INDUS CHAS NOX RM Cluster 0 0.2451 11.24 6.21 0.06 0.47 6.35 1 3.4827 3.84 14.91 0.08 0.62 6.01 2 0.7215 8.62 9.73 0.07 0.54 6.72 AGE DIS RAD TAX PTRATIO Cluster 0 52.41 4.71 4.82 282.41 17.82 1 78.35 2.94 9.21 449.73 18.91 2 63.27 3.87 5.74 329.18 17.24 B LSTAT Cluster 0 385.24 8.12 1 355.71 16.84 2 389.63 7.91 Final Output: Clustered dataset saved successfully.

How to Understand the Output

Cluster Counts shows how many housing records were assigned to each cluster.

Cluster Summary displays the mean value of each selected feature within each cluster. This allows you to compare the characteristics of the groups.

PCA Visualization reduces the 13-dimensional feature space to two dimensions so that the cluster assignments can be displayed on a scatter plot.

The cluster labels such as 0, 1 and 2 are simply identifiers assigned by the algorithm. They do not inherently indicate a ranking or quality level.

⬇ View Boston Housing Dataset on Kaggle

Created by Bijan Krishna Paul