Machine Learning Made Easy With PyCaret • Moez Ali • GOTO 2022

Machine Learning Made Easy with PyCaret

• Comparing models shows the average results, not individual fold scores
• Log all metrics, including plots, during experimentation
• Use Docker for reproducibility and version control
• Pre-processing is important; PyCaret includes many methods
• Feature importance is used for explanation
• CatBoost, LightGBM, and SHAP are integrated libraries
• Low code environment is a key differentiator
• Citizen data scientists are halfway between business users and data scientists
• PyCaret has a modular design, with many building blocks
• Tensus notebooks for data exploration and visualization
• Fast API is used for building APIs
• PyCaret has integration with Fugue for distributed computing
• Random forests and gradient boosting are important algorithms
• Explanability is a key concept in machine learning
• PyCaret is open-source and has a large community
• It’s a good tool for small and medium-sized companies