Customer Segmentation - Advanced Analysis
Interactive demo showcasing customer clustering analysis with K-Means. Explore data stories, clustering patterns, and predict segments for new customers.
Data Overview Analysis
3,921
Tổng số khách hàng
18,021
Tổng số giao dịch
£18.60
Doanh thu trung bình/giao dịch
Revenue Over Time
From Date
To Date
Shopping Behavior by Hour and Day
Insights:
- Heatmap shows shopping patterns by hour (0-23) and day of week
- Revenue Over Time shows overall sales trend (12 months)
- Filter by date range to zoom into peak months (Christmas, etc.)
Explore K-Means Clustering Algorithm
Adjust the slider to select different numbers of clusters (K) and see how the algorithm divides customers into different groups.
Determine Optimal Number of Clusters
Explanation:
- Elbow Method: Find the "elbow" point where increasing K doesn't significantly reduce inertia
- Silhouette Score: Higher is better. Clusters are more distinct when score is high
- Recommendation: K=3 or K=4 are both good choices
Visualize Clusters in PCA Space
How to Use:
- Each point represents one customer
- Color indicates which cluster the customer belongs to
- When changing K, clusters will be instantly updated from cache
Deep Analysis: Characteristics of Each Cluster
Select a cluster to see detailed characteristics of customers in it. The Radar chart shows how this cluster differs from the overall average.
Radar Chart - Comparison with Overall Average
How to Read Radar Chart:
- Each axis = 1 customer characteristic (normalized 0-1 scale)
- Further from center = higher value for that characteristic
- Shape of polygon represents the cluster's profile
- Compare clusters by looking at shape and size
Project: Advanced Customer Segmentation
Data: Online Retail (2010-2011) - Customers: 3,920+ - Transactions: 354,000+
Built from: Project by Dr.Nguyen Thai Ha