Project 2 - Customer Churn Prediction

Customer churn occurs when customers stop using a service. Businesses use churn analysis to understand customer behavior and identify customers who may be at risk of leaving.

This project introduces a binary classification workflow using Scikit-learn.

Assume a prepared dataset named customers.csv contains numerical features and a target column called Churn, where 1 indicates that a customer left and 0 indicates that the customer stayed.

Step 1: Prepare the Data

import pandas as pd

from sklearn.model_selection import train_test_split

df = pd.read_csv("customers.csv")

X = df.drop(columns=["Churn"])

y = df["Churn"]

X_train, X_test, y_train, y_test = train_test_split(

X,

y,

test_size=0.2,

random_state=42,

stratify=y

)

The example assumes that the input features are already numeric and contain no missing values. If the dataset contains categorical columns or missing values, add suitable preprocessing before training.

Step 2: Train a Classification Model

from sklearn.pipeline import Pipeline

from sklearn.impute import SimpleImputer

from sklearn.preprocessing import StandardScaler

from sklearn.linear_model import LogisticRegression

model = Pipeline([

("imputer", SimpleImputer(strategy="median")),

("scaler", StandardScaler()),

("classifier", LogisticRegression(max_iter=1000))

])

model.fit(X_train, y_train)

predictions = model.predict(X_test)

The pipeline fills missing numerical values with the median, standardizes features, and trains a logistic regression classifier. For mixed data types, use a ColumnTransformer to preprocess numerical and categorical columns separately.

Step 3: Evaluate the Model

from sklearn.metrics import (

classification_report,

confusion_matrix,

roc_auc_score

)

probabilities = model.predict_proba(X_test)[:, 1]

print(classification_report(y_test, predictions))

print(confusion_matrix(y_test, predictions))

print("ROC-AUC:", roc_auc_score(y_test, probabilities))

Accuracy alone may hide poor performance when churn is uncommon. Precision and recall help evaluate the model's ability to identify customers who are likely to leave.

Possible project outcomes:

• Identify customer groups with elevated churn risk.
• Explore factors associated with customer departures.
• Help business teams investigate retention opportunities.
• Compare model performance at different classification thresholds.

A churn prediction is not proof that a customer will leave. The model should support further investigation rather than automatically determine how customers are treated.