Machine Learning Made Easy With PyCaret
summary
PyCaret claims to replace hundreds of lines with a few by wrapping scikit-learn, XGBoost, LightGBM, CatBoost, spaCy, Optuna, Hyperopt and Ray and by default imputes missing values, one‑hot encodes and splits train/test. Demo runs compare_models (10‑fold CV across many estimors) producing a gradient boosting MAE ≈2701, tuning drops a decision tree from 3383 to ~2000 and holdout MAE is 2392. It logs to MLflow, exports scikit‑learn pipelines, uploads to AWS S3, spins up FastAPI + Docker, and can transpile models to Java/PHP (trees produce huge code).
AI-written summary · about
chapters (11)
- 0:00Intro
- 2:51AI & data landscape – 2022
- 3:37PyCaret
- 7:02Phases of ML
- 8:47ML life cycle
- 11:12Data prep, model training & selection
- 11:42Features
- 13:07Use-cases supported in PyCaret
- 15:02Integrations
- 16:51Demo
- 43:17Outro
related talks
57mGood Practices for Applied Machine Learning
31mPut Your Machine Learning on Autopilot- 57mGood Practices for Applied Machine Learning
01h 24mInpsect and try to interpret your scikit-learn machine-learning models
27mPyTesting the Limits of Machine Learning
38mFantastic ML apps and how to build them
35mScikit-Learn can do THAT?!
37mMachine Learning, The Right Way