Unsupervised Learning: Association and
Clustering
Unsupervised learning is a powerful tool in machine learning that allows us to uncover hidden patterns and structures within data. It differs from
supervised learning by not requiring labeled data for training.
[Link]
Introduction to Unsupervised Learning
Exploring Data Discovering Structure Unveiling Anomalies
Unsupervised learning explores data It identifies underlying relationships Unsupervised learning helps detect
without prior knowledge, aiming to and groupings within data, revealing unusual data points that deviate from
discover inherent patterns and its inherent organization. the norm, highlighting potential
insights. outliers.
[Link]
Importance of Unsupervised Learning
Data Exploration Customer Segmentation Anomaly Detection
It allows us to discover hidden patterns, Grouping customers into distinct segments Unsupervised learning can detect unusual
relationships, and anomalies in data, based on their behaviors and preferences patterns in data, such as fraudulent
providing a deeper understanding of its helps tailor marketing strategies. transactions or equipment malfunctions.
structure.
[Link]
Association Rule Mining: Concept and Applications
1 Identifying Relationships
Association rule mining discovers interesting relationships between items in a dataset.
2 Market Basket Analysis
It helps understand customer buying patterns and suggest products based on their past purchases.
3 Medical Diagnosis
Analyzing medical records can reveal associations between symptoms and diseases, aiding diagnosis.
4 Fraud Detection
Identifying unusual patterns in financial transactions can help detect fraudulent activities.
[Link]
Apriori Algorithm for Association Rule Mining
• Candidate Generation
The algorithm generates candidate itemsets by combining frequent itemsets from the previous iteration.
• Support Calculation
It calculates the support of each candidate itemset, which is the frequency of its occurrence in the dataset.
• Confidence Calculation
The confidence of a rule is calculated by dividing the support of the rule by the support of its antecedent.
[Link]
Clustering Algorithms: K-Means
and Hierarchical Clustering
K-Means Clustering Hierarchical Clustering
It partitions data points into k It creates a hierarchy of clusters,
clusters, where each point belongs starting with individual data points
to the cluster with the nearest and merging them based on their
mean. similarity.
[Link]
Choosing the Optimal Number of Clusters
Elbow Method Plots the within-cluster sum of squares (WCSS) against the number
of clusters. The optimal number of clusters is where the WCSS
starts to decrease less rapidly.
Silhouette Score Measures how similar a data point is to its own cluster compared to
other clusters. A high silhouette score indicates good clustering.
[Link]
Evaluating Clustering Performance
Silhouette Score Davies-Bouldin Index Calinski-Harabasz Index
Measures how well a data point fits within its Evaluates the ratio of within-cluster distances Measures the ratio of between-cluster
assigned cluster. to between-cluster distances. variance to within-cluster variance.
[Link]
Real-World Applications of Unsupervised Learning
Customer Segmentation Anomaly Detection
Unsupervised learning helps group customers into distinct segments, It can identify unusual patterns in data, such as fraudulent
allowing businesses to tailor marketing strategies. transactions or equipment failures, enabling early detection
and prevention.
Market Basket Analysis
Dimensionality Reduction
In retail and e-commerce, unsupervised learning can be used for Applied in scenarios where large datasets need to be visualized or
discovering associations between products. processed efficiently, such as in genetics, finance, or text analysis.
Recommender Systems Social Network Analysis
Used to find communities within social networks, detect
influencers, and analyze social behaviors.
Content-based filtering in recommender systems that do not rely on
explicit feedback.
[Link]
THANKS
[Link]
