Machine Learning 101 — Classification vs. Clustering

Part of a Series on Machine Learning Concepts

Determination of age group based on physical characteristics

Classification and Clustering are two types of learning methods that attempts to categorize records based on one or more features in the data. While there are similarities in their objectives, their approaches are different.

Classification

Classification uses supervised learning techniques to find the relationship between the feature(s) and the assigned label(s). Upon training with sufficient data samples, the resulting model can be used to predict the label (and probability) that a given data point resembles.

Algorithms:

• Logistic Regression
• Decision Tree/Random Forest
• Neural Networks
• Naive Bayes
• K-Nearest Neighbors
• Support Vector Machines

Common Applications:

• Risk Assessment Model
• Spam Detection
• Fraud Detection

Clustering

Clustering uses unsupervised learning techniques to organize unlabeled data points into homogeneous groups where those within a “cluster” have similarities and “clusters” have dissimilarities. Since data is not pre-labeled, there is no training process. Instead, clusters are created using similarity functions that measure the distance between points. Relative to classification, clustering is considered a less complex approach and well suited for large datasets.

Algorithms:

• K-means
• K-medoids
• Density Based
• Hierarchical

Common Applications:

• Recommender System
• Customer Segmentation

Comparison Chart

Comparison chart between classification and clustering