Project: Assessing Customer Churn Using Machine Learning

The telecommunications (telecom) sector in India is rapidly changing, with more and more telecom businesses being created and many customers deciding to switch between providers. "Churn" refers to the process where customers or subscribers stop using a company's services or products. Understanding the factors that influence keeping a customer as a client in predicting churn is crucial for telecom companies to enhance their service quality and customer satisfaction. As the data scientist on this project, you aim to explore the intricate dynamics of customer behavior and demographics in the Indian telecom sector in predicting customer churn, utilizing two comprehensive datasets from four major telecom partners: Airtel, Reliance Jio, Vodafone, and BSNL:

telecom_demographics.csv contains information related to Indian customer demographics:

Variable	Description
`customer_id`	Unique identifier for each customer.
`telecom_partner`	The telecom partner associated with the customer.
`gender`	The gender of the customer.
`age`	The age of the customer.
`state`	The Indian state in which the customer is located.
`city`	The city in which the customer is located.
`pincode`	The pincode of the customer's location.
`registration_event`	When the customer registered with the telecom partner.
`num_dependents`	The number of dependents (e.g., children) the customer has.
`estimated_salary`	The customer's estimated salary.

telecom_usage contains information about the usage patterns of Indian customers:

Variable	Description
`customer_id`	Unique identifier for each customer.
`calls_made`	The number of calls made by the customer.
`sms_sent`	The number of SMS messages sent by the customer.
`data_used`	The amount of data used by the customer.
`churn`	Binary variable indicating whether the customer has churned or not (1 = churned, 0 = not churned).

# Import libraries and methods/functions
import pandas as pd
from sklearn.preprocessing import StandardScaler, OneHotEncoder 
from sklearn.model_selection import train_test_split
from sklearn.linear_model import RidgeClassifier, LogisticRegression  # <-- Added LogisticRegression import
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, confusion_matrix

# Start your code here!
# Load datasets
telco_demo = pd.read_csv("telecom_demographics.csv")
telco_usage = pd.read_csv("telecom_usage.csv")

# Quick look
print(telco_demo.head())
print(telco_usage.head())

# Merge on customer_id
churn_df = pd.merge(telco_demo, telco_usage, on="customer_id", how="inner")
print(churn_df.head())

# Calculate churn proportion
churn_counts = churn_df['churn'].value_counts()
churn_rate = churn_counts[1] / len(churn_df)
print("proportion of churn:", churn_rate)

# Inspect column info
churn_df.info()

# One Hot Encoding for categorical variables

churn_df = pd.get_dummies(churn_df, columns=['telecom_partner', 'gender', 'state', 'city', 'registration_event'])

print(churn_df.head())

# Initialize scaler 
scaler = StandardScaler() 


# 'customer_id' is not a feature
features = churn_df.drop(['customer_id', 'churn'], axis=1)
features_scaled = scaler.fit_transform(features)

# Target variable
target = churn_df['churn']

# Perform 80-20 split
X_train, X_test, y_train, y_test = train_test_split(features_scaled,target, test_size=0.2, random_state=42)

print("Training set size:", X_train.shape[0]) 
print("Testing set size:", X_test.shape[0])

# Instantiate the Logistic Regression
logreg = LogisticRegression(random_state=42)
logreg.fit(X_train, y_train)

# Logistic Regression predictions
logreg_pred = logreg.predict(X_test)

# Logistic Regression evaluation
print(confusion_matrix(y_test, logreg_pred))
print(classification_report(y_test, logreg_pred))

# Instantiate the Random Forest model
rf = RandomForestClassifier(random_state=42)
rf.fit(X_train, y_train)

# Random Forest predictions
rf_pred = rf.predict(X_test)

# Random Forest evaluation
print(confusion_matrix(y_test, rf_pred))
print(classification_report(y_test, rf_pred))

# Which accuracy score is higher? Ridge or RandomForest
higher_accuracy = "RandomForest"