From: Customer churn prediction in telecom using machine learning in big data platform
Feature name | Feature description |
---|---|
In-Degree | Number of friends connecting with the customer |
Out-Degree | Number of friends the customer connecting with |
Max-Cosine-Sim-MTN | Maximum Cosine similarity with other operators’ customers |
Max-Cosine-Sim-SyriaTel | Maximum Cosine similarity with SyriaTel customers |
Max-Jaccard-SIM-MTN | Maximum Jaccard similarity with other operators’ customers |
Max-Jaccard-Sim-SyriaTel | Maximum Jaccard similarity with SyriaTel customers |
SR | Weighted Sender Rank in social graph |
PR | Weighted Page Rank in social graph |
PF “Social Power Factor” | Average of weighted Page Rank and Sender Rank in social graph |
Betweenness | # of short paths between any two people in the social network passes through this customer node |
LLC “Local Cluster Coefficient” | How much the customer friends know each other |
NC “Neighborhood Connectivity” | The number of friends and friends of friends for the customer |