Classifying the variety of customers online engagement for churn prediction
summary
Proposes a mixed penalty logistic regression that blends absolute (median) and squared (mean) regularization to enforce sparsity and avoid overfitting on high-dimensional CRM data from a European gaming operator. Tunes the penalty weight lambda and uses bootstrapped robustness checks to handle class imbalance, reporting clearer feature selection and boosted recall for churners versus unpenalized logistic models and standard lasso/ridge, with a claimed computational advantage for optimization.
AI-written summary · about
chapters (4)
- 0:00Start
- 9:55Other Applications of Machine Learning
- 22:23Regularization Model
- 24:53Main Steps of the Proposed Algorithm
related talks
24mUnlocking revenue: newcomer classification for boosted profits
27mPreventing churn like a bandit
36mCustomer Lifetime Value: Models, Metrics and a Multitude of Uses
30mDetecting Hidden Bias in Datasets Before Models Fail
35mUsing Survival Analysis to understand customer retention
38mOnline customer targeting as a classification problem
30mWinning (Some) Churners Back
31mHow to Validate you Client Churn Model