NadjibSb / PFE-Segment-specefic-Modeling

Graduation project | An approach to build better predictive models.

Segment-specific Modeling

An approach to build better predictive models.
Use Case : Telecom

1- Clustering

Cluster clients according to their behavior (Calls, Data, SMS, Recharge)

Development steps :

• Data exploration
• Data cleaning
• Modeling : using k-means & aggmolorative clustering
• Model evaluation
• Visualization
• Deploy the model via REST API using flask

2- Churn prediction

Build a predictive model for each cluster of client instead of one model for all the clients Use Case : Churn prediction

Development steps :

• Data exploration
• Segment-specific Modeling: using Auto-ML with TPOT
• Models evaluation
• Deploy the models via REST API using flask

3- MAPE-K Control Loop

Self-adaptive system based on MAPE-K architecture to monitor the resulted models

Components :

• Monitor
• Analyze
• Plan
• Execute
• Knowledge