Custom-made games with machine learning and big data
summary
Yokozuna Data describes a production ML platform that predicts churn 2–6 months ahead and individual LTV, combines churn models with CNNs for robust lifetime spend forecasts, and uses unsupervised embeddings plus collaborative filtering for per-player item recommendations. Pipeline runs daily on Spark/Cassandra at millions MAU, supports A/B testing and event sequence simulation, and claims contest wins at the IEEE 2017 game data mining challenge. Example business impacts shown include +$1.3M from a 10% VIP uplift and a $4M/month lift from converting potential payers.
AI-written summary · about
chapters (5)
- 0:00Start
- 6:05Item Recommendation
- 14:44Deep Learning
- 19:20Dashboard
- 21:43Simulation of Game Events and Marketing Campaigns
related talks
01h 00mHow to win at monetizing games as a service
19mMachine Learning is a new age of Personalized Game Monetization
27mBig Data in Mobile Games
20mUsers acquired - Now what?: Maximising revenue from existing players
36mHow to Enhance the Gaming Experience Through Content Micro-personalisation
24mEngaging new and diverse audiences with ML
23mThe unreasonable effectiveness of data for better monetisation
33mProduction ML Models in the Mobile Gaming Industry