Customer_Segmentation: Projec Overview

---

• Create a tool that runs monthly to categorize each customer_id(msisdn)
• The measures used to categorize these customers are: Revenue, Recency & Frequency aka RFM
• The aim is to have 3 categories: High Value Customers, Low Value Customers, Mid Value Customers
• Present to marketing team to target specific customers for specific campaigns

Code and Resources Used

---

Python Version: 3.7 Packages: pandas, numpy, sklearn, KMeans, matplotlib, seaborn

Data

---

• MSISDN object
• LAST_RECHARGE_DATE datetime64[ns]
• BTS_MU_CITY object
• BTS_MU_LAT object
• BTS_MU_LON object
• COUNTRY object
• TARIFF_TYPE object
• ASPU float64
• RCH_COUNT_VOUCHER float64
• REGION_CLUSTER object
• TERRITORY object

EDA

---

• Recency = Last day of month - max date of recharge per msisdn

• Frequency = Count of number of recharges in given month per msisdn

• Revenue = Sum ASPU per msisdn

Model Development

---

unsupervised count optimal number of clusters:

• Allocate cluster group to all customers for each RFM
• Create new feature 'Overall Score' - sum eaach customer's score across Revenue, Frequency and Recency cluster

• Group each overall score into value bracket(High Value Customers, Low Value Customers, Mid Value Customers)
• show graphical characteristics of each value group

Frequency_Recency_Scatter

Revenue_Recency_Scatter

Revenue_Frequency_Scatter