🎯 iGaming Retention Test - Part 4

Predicting Churn with Logistic Regression

Now that we’ve confirmed retention uplift, let’s try to predict churn to understand which factors drive player loss.

Main Goals of the day

• Train a Logistic Regression model to predict churn
• Use sessions, deposits, and feature usage as predictors
• Evaluate accuracy, recall, and precision
• Interpret feature coefficients

Step by Step

📍 Step 1: Encoded features and target (churn)
📍 Step 2: Split dataset into train/test
📍 Step 3: Fitted logistic regression with sklearn
📍 Step 4: Evaluated model performance metrics

Insights

The model performs slightly above random:
AUC = 0.547, Accuracy = 54%, Recall = 52%

Early churn models often struggle due to class imbalance — still, it reveals interesting patterns.

Code Snippet

from sklearn.linear_model import LogisticRegression
from sklearn.metrics import roc_auc_score

X = df[['sessions', 'deposits', 'feature_used']]
y = df['churn']

model = LogisticRegression()
model.fit(X, y)
auc = roc_auc_score(y, model.predict_proba(X)[:,1])
print("AUC:", auc)

Next Step

Add SQL-based aggregation to structure metrics and retention KPIs

Enrich the dataset with session-level metrics before retraining