Skip to content

Welcome to Pilz

Pilz is a machine learning library that makes classification easy to understand and deploy. Named after the German word for mushroom/fungus - organisms that are neither plant nor animal - Pilz bridges the gap between traditional decision trees and neural networks.

Why Pilz?

Traditional machine learning models are often "black boxes" - you put data in, get predictions out, but understanding why is difficult. Pilz changes this by generating readable SQL rules that you can directly deploy to your database.

Key Features

Feature Description
Three-Way Splits Trees have Left, Neutral, and Right branches for better handling of ambiguous cases
SQL Output Generates actual SQL that runs directly in DuckDB
Interpretable Every decision is a clear rule you can read
Multi-Class Handles any number of classes
Feature Magic Automatic categorization of continuous and categorical features

Quick Example

# Train a model
pilz train --datacard mydata.yaml --trainsettings settings.yaml

# Get predictions as SQL
pilz eval --datacard mydata.yaml --evalsettings settings.yaml

The output includes ROC curves, accuracy metrics, and deployable SQL rules.

What You'll Learn

  1. Getting Started - Install and run your first model in 5 minutes
  2. Tutorials - Step-by-step guides with real datasets
  3. Core Concepts - How the algorithm actually works
  4. Reference - Complete command and settings reference

Ready to Start?

Installation Guide


Pilz - Making ML decisions transparent and deployable.