Business Case Analysis

About the Dataset

The dataset contains customer-level information for four consecutive months - June, July, August and September. The business objective is to predict the churn in the last (i.e., the ninth) month using the data (features) from the first three months.

Data Preparation

1. Derive new features to filter high-value customers

New Feature = Data greater than or equal to 70th percentile of the average recharge amount in June and July.

2. Tag churners and remove attributes of the churn phase.

Identifying churned high-value customers based on the last month(September) 1=Churned Customers 0=Non-Churned Customers

Model Building

• Preprocess data Exploratory
• Data Analysis
• Derive new features.
• Reduce the number of variables using PCA Handle class imbalance
• Train different models and perform appropriate hyperparameters
• Evaluate the models using appropriate evaluation metrics
• Choose a model based on some evaluation metric

Recommendations to Business based on Findings

We can infer from our analysis that certain features are high indicators of whether a customer will churn or not. For Telecom to reduce the number of churn and maintain these high-valued customers, it has to take note of these features and monitor them consistently to identify the stage of those customers that are likely to churn and address their pain-point as soon as possible to prevent them from churning. Another key factor to be considered is that if the recharge amount of the customer starts to reduce during the month when the customer has shown consistent activeness, then it's a sign of churn and must be addressed as soon as possible. Age on network is also a key indicator for identifying the churn. If the age is less than 500 days and their usage is reduced, then the customer is likely to churned. The other key indicator the company should consider is data usage & amount.