Machine Learning for Gamers - Dungeon Forecasts & Dragon Regressions
summary
Shows concrete recipes: one‑hot encoding for dragon color and categorical race, one‑hot encoding, numeric regression to predict a dragon hoard from age, color and hit points, and decision‑tree/forest classification that scores classes from six ability scores and race. Uses 80/20 train/test splits, watches overfitting, and runs anomaly detection on characters and server metrics. Demonstrates forecasting encounters per session (example predicted 5.03 vs actual 3 on Jan 1) and an impact analysis that finds switching from Pathfinder to Fifth Edition slows sessions.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:33Who is this Guy?
- 2:11Who is this Company?
- 11:23Target & Features
- 12:01Encoding & Imputation
- 16:31Algorithms
- 16:57Train & Test
- 21:24About Dragons
- 26:03About Characters
- 27:55Asking the Question
- 35:27About Gaming Sessions
- 36:54Training Data
related talks
48mMachine learning the hard way -- a story about ponies
23mCustom-made games with machine learning and big data
48mMachine Learning for Boardgame AI in Race for the Galaxy
06mPredicting Game Difficulty and Churn Without Players
22mPredict the future (or don't) with machine learning in applications
58mUnleashing the Power of AI to Personalize Games
37mHow I trained an AI Model to Beat the 1990's Arcade Game Double Dragon
31mBecoming a Pokémon Master with DVC: reproducible machine learning experiments