What Is Clustering?

Written by Coursera Staff • Updated on

Professionals across industries use cluster analysis to explore data and inform decision-making. Learn more about different types of clustering, why cluster analysis is important, and techniques to visualize your clusters.

[Featured Image] A data scientist sits at a desk with a laptop and relies on clustering to gain insights.

Clustering is a technique used in data analysis to organize data into clusters based on similar features. The idea is that similar data are in each cluster, showing natural grouping within the data. You can choose to cluster based on different types of attributes like color, size, or type. Cluster analysis is a form of unsupervised learning, meaning it doesn’t rely on predefined categories or labels and instead discovers inherent groups in the data.

In this article, we will explore what clustering is, why it is important, how you might use this method, and examples of different types of cluster analyses. 

What is data mining?

Data mining involves finding patterns, trends, and information in large volumes of data. It uses algorithms and statistical methods to uncover relationships and insights within data that may not be immediately obvious. Cluster analyses are one type of data mining algorithm used to uncover data characteristics through a natural grouping of the information.

Why is clustering important?

Clustering is important in data analysis for several reasons, including identifying patterns and structures within large data sets that may not be immediately obvious. By organizing data into clusters, analysts can more easily interpret and understand the data, leading to more informed decision-making.

Clustering is important across various fields and applications, helping professionals explore their data and identify directions for further analysis. For example, in business and marketing, companies can segment their customers into groups they can use for targeted marketing strategies, helping them optimize resources and enhance customer satisfaction. By clustering customers based on purchasing behavior, a business can decide how to market to each group most effectively.

What is clustering used for?

Professionals use clustering methods in a wide variety of industries to group data and inform decision-making. Some ways you might see clustering applied include the following:

  • Business: Companies use clustering for customer segmentation, which means grouping customers based on their behavior and characteristics.

  • Machine learning: Clustering can organize large data sets and improve model performance.

  • Ecology: Clustering can classify plants or animals based on genetic or physical characteristics, aiding in biodiversity studies and conservation efforts.

  • Social networking: Clustering helps identify communities within social networks by looking at characteristics and relationships.

  • Investment: Clustering can inform stock price trends and investment algorithms, improving financial returns.

  • Finance: Financial institutions cluster transactions to detect fraudulent activities, often hidden from common detection methods.

  • Climate analysis: Cluster analysis can identify weather trends and patterns, informing scientists on metrics such as atmospheric pressure.

  • Resource allocation: Companies can use cluster analysis to identify areas that require more attention, such as needing more personnel or certain types of resources.

Benefits and drawbacks of clustering

Choosing cluster analyses for your data can offer many benefits. Some advantages you might experience include:

  • Improved understanding of your data

  • Doesn’t rely on previous knowledge of data features

  • Several methods suited for different applications

  • Informed decision-making

  • Diverse applications across various industries

When considering advantages, it’s also important to consider disadvantages. Limitations to be aware of include:

  • Not able to make predictions

  • Difficulty with clusters of different sizes and densities with some methods

  • Sensitive to outliers

Different types of clustering

Understanding how each method works can help you decide which is right for your data when choosing a clustering algorithm. While methods differ, each algorithm has the same goal: to classify data into similar groups.

Hierarchical clustering

Hierarchical clustering is a clustering method that methodically groups data, either from a top-down or bottom-up approach, known as divisive and agglomerative hierarchical clustering, respectively. For divisive clustering, the tree's top point (root) includes all of the data. The data then branches into large subgroups, which branch into smaller subgroups, and so on. For example, you might start with a group of animals, which is then classified as mammal, reptile, and so on, before being further classified into species. For agglomerative clustering, the reverse process applies. Every animal is classified individually first, then categorized into small groups before these groups join to form larger classifications until all data is in one group.

K-means clustering 

K-means clustering algorithms work by clustering data into a pre-defined number of groups. This method works by placing the pre-defined number of centroid values. The goal is to minimize the distance between the data point and the nearest centroid value. The data iteratively groups and the centroid values adjust based on the data distribution until the “minimum distance” is found so the clusters best represent the underlying organization. In most cases, data points can only belong to one cluster. However, “fuzzy” k-means algorithms allow the inclusion of data points in several clusters.

Mixture-based clustering

Mixture-based clustering models, such as Gaussian mixture models (GMM), reference different probability distributions to classify where data points belong. Each cluster corresponds to a distribution, and data points group based on the likelihood of belonging to those distributions. This method is flexible and can accommodate clusters of different sizes and shapes.

Visualization tools in cluster analysis

Both heat maps and self-organizing maps (SOMs) can enrich your cluster analysis by offering distinct ways to visualize and interpret your data. While heat maps provide a direct visual representation of the structure of your data with color gradients, SOMs show data similarities through proximity between points. 

Heat maps

Heat maps in cluster analysis are graphical representations of data where a color shows each value. They are particularly useful if you’re visualizing the presence or magnitude of phenomena, allowing you to quickly see patterns, correlations, and trends in your data. In a heat map, closely related data points have similar colors, indicating they belong to the same cluster. This visualization can help you see underlying structures in the data before using formal clustering methods and is commonly used in fields such as biology and genetics. 

Self-organizing maps

Self-organizing maps are an artificial neural network that takes high-dimensional data and outputs a two-dimensional representation. You might see this type of representation used in business applications, bioinformatics, and data mining. SOMs reveal clusters and relationships that might not be apparent from traditional clustering techniques. By mapping high-dimensional data onto a two-dimensional grid, you can use SOMs to understand complex data better, helping you identify patterns and explore data relationships.

Getting started with Coursera

Learn more about cluster analysis with exciting courses on the Coursera learning platform. You can choose between several courses depending on your field and skill level, including High-Dimensional Data Visualization Techniques Using Python, Business Analytics for Decision Making, or Statistics for Marketing.

Keep reading

Updated on
Written by:

Editorial Team

Coursera’s editorial team is comprised of highly experienced professional editors, writers, and fact...

This content has been made available for informational purposes only. Learners are advised to conduct additional research to ensure that courses and other credentials pursued meet their personal, professional, and financial goals.