Machine Learning and Online Experiments to Prevent Churn on GloboPlay
summary
Describes using machine learning at GloboPlay to reduce subscriber churn by identifying users most likely to cancel via clustering and classification on imbalanced data, and by analyzing correlated metrics. Covers recommendation algorithms and running simultaneous A/B tests with methods to prevent experiments from interfering with each other, plus practical guidance on techniques and how to apply data science to produce measurable product results.
AI-written summary · about
chapters (11)
- 0:00Project introduction
- 3:31TIHany metric challenges
- 7:10Handling imbalanced data
- 10:20Modeling and testing process
- 14:35Factors influencing churn
- 17:04Globoplay focus shift
- 19:10Title prediction for retention
- 23:22Personalized recommendations
- 27:56Business metrics and experiments
- 31:49Experiment isolation
- 35:18Results and key takeaways
related talks
23mCustom-made games with machine learning and big data
27mUnlocking Product Growth With Big Data, Data Science And Machine Learning
31mAppDev with Machine Learning and AutoML
41mRecommendations for Building Machine Learning Software
13mOptimizing Subscription Services with Uplift Modeling
49mData Science in Gaming
27mPreventing churn like a bandit
24mMachine Learning for Analytics Team