Customer Segmentation using Clustering - K-Means, DBSCAN, Hierarchical Clustering
MY ROLES
TIMELINE
2019
TOOLS USED
Clustering, K-Means, DBSCAN, Python, Pandas, Scikit-learn, Data Visualization, Customer Analytics, Marketing Analysis, Silhouette Analysis, Elbow Method
DESCRIPTION
This project applied various clustering algorithms to segment customers based on their responses to marketing campaigns and transaction behavior. Using data from email newsletters and purchase history, the analysis compared multiple clustering techniques (K-Means, DBSCAN, hierarchical clustering) to identify natural groupings in customer behavior that could inform targeted marketing strategies.
OUTCOMES
- Identified optimal customer segments using multiple clustering algorithms with DBSCAN yielding the highest silhouette score
- Discovered 2-4 distinct customer behavior patterns that could be leveraged for targeted marketing
- Created visualization framework for comparing clustering algorithm performance
FEATURES
Multi-algorithm comparison (K-Means, DBSCAN, hierarchical), silhouette score analysis, elbow method implementation, gap statistic calculation, customer segment profiling
CHALLENGES
Determining the optimal number of clusters without labeled data. Handling mixed data types from marketing and transaction sources. Interpreting clusters in a meaningful business context that could drive actionable marketing strategies.
APPROACH
Preprocessed marketing newsletter and transaction data from Excel workbooks. Implemented and compared multiple clustering techniques including K-Means and DBSCAN. Evaluated models using silhouette scores, elbow method, and gap statistics. Identified DBSCAN as the superior method based on silhouette scores. Recommended focusing on 2-4 customer segments that showed the clearest separation.