Basic Customer Segmentation

This workflow implements a basic customer segmentation through a clustering procedure. The data is of a telecommunication company; its customer records. Feature selection has been done in an another workflow by drawing boxplots and densityplots. Those fearures have been selected which show good relationship with the target, churn. Eight features are considered. Weka widget is used for clustering. Three clusters can be clearly seen in the 2D scatterplot as also 3D scatterplot.
To normalize any one of the three techniques can be considered. Clusters remain unaffected. Also distance measure can be either Euclidean or Manhattan but again the three clusters remain unaffected.
Simple workflow to draw screeplot is also shown.

External resources

• Notes on Kmeans algorithm on Moodle
• Recent Advances in Clustering--A Survey
• Proximity Measures
• Cluster Analysis--Basics
• Cluster Analysis--Wikipedia