pycaret/pycaret

Features

• Machine Learning Workflow Libraries - Executes machine learning workflows using a standardized library of scikit-learn estimators and pipeline objects.
• Automated Machine Learning - Automates model selection, hyperparameter tuning, and pipeline construction to find the best performing model.
• AI Agent Integrations - Analyzes datasets and debugs failures using large language models to suggest retraining strategies based on drift reports.
• AI-Powered Data Assistants - Uses AI-powered assistants to analyze datasets and suggest optimal experiment configurations.
• Training Execution - Executes machine learning training tasks across local thread pools or remote compute clusters.
• Model Performance Benchmarks - Iterates through a model registry to train multiple candidates and ranks them based on performance metrics.
• AI-Powered Dataset Analyzers - Analyzes uploaded files using LLMs to suggest target columns, task types, and preprocessing strategies.
• ML Experiment Copilots - Provides an AI assistant that analyzes datasets, designs experiments, and explains model results.
• Developer AI Assistance - Consults on dataset preparation and experiment design using integrated large language model copilots.
• End-to-End Training Pipelines - Wraps model comparison and hyperparameter tuning into a single routine that returns an optimized pipeline.
• Experiment Tracking - Includes tools for logging, versioning, and visualizing the lifecycle of machine learning experiments.
• LLM Gateways - Routes natural language requests through a unified interface to external large language models for experiment guidance.
• LLM Provider Integrations - Routes requests through a provider gateway to receive advisory suggestions for experiment design and drift analysis.
• Machine Learning Model Registries - Implements a centralized catalog for storing and versioning trained pipelines for production promotion.
• Low-Code Machine Learning Tools - Provides a visual interface and abstraction layer for building ML models with minimal coding.
• Automated ML (AutoML) - Implements AutoML capabilities to automatically select and tune the best models based on dataset statistics.
• Serving Endpoints - Publishes registered models to serving endpoints that manage how they are loaded for real-time inference requests.
• Hyperparameter Tuning - Optimizes model performance through iterative search of hyperparameter grids and distributions.
• MLOps Platforms - Provides a comprehensive platform for managing the operational lifecycle of AI models.
• Model Deployment Pipelines - Implements deployment pipelines to promote fitted models to a serving registry and production prediction APIs.
• Model Predictions - Generates predictions for classification, regression, clustering, or anomaly tasks using trained estimators.
• Model Performance Selection - Automatically identifies the best performing machine learning algorithm for a specific analytical task.
• Model Serving APIs - Exposes trained machine learning models as network-accessible REST APIs for external application integration.
• Model Training Pipelines - Builds integrated pipelines that combine data preprocessing and trained estimators from a registry.
• Model Versioning - Tracks trained pipelines in a central catalog to enable versioning, promotion, and rollback of production artifacts.
• Model Versioning Systems - Tracks iterations of machine learning models and their associated artifacts in a central registry.
• Time Series Forecasting - Analyzes temporal patterns in data to predict future values using specialized time-series modeling techniques.
• AI Agents and Assistants - Uses natural language agents to analyze datasets, design experiment configurations, and explain model results.
• Pipeline Optimizers - Iterates through preprocessing variants and model candidates to automatically identify and fit the highest-performing pipeline.
• Experiment Management - Provides tools for bootstrapping projects and configuring experiments via dynamic, parameter-driven forms.
• MLOps and Deployment - Manages the full machine learning lifecycle from experiment tracking to production monitoring.
• ML Workflow - Provides a visual interface to set up datasets and experiments, eliminating the need for manual configuration files.
• AI Workflow Designers - Uses LLM-powered copilots to guide users through dataset consultation and the design of machine learning experiments.
• MLOps Pipeline Automation - Automates the transition of ML models from experimental development to production-ready scalable pipelines.
• ML Lifecycle Orchestration - Organizes data, experiments, and runs into isolated projects to manage the end-to-end machine learning lifecycle.
• ML Model Hosting - Provides infrastructure for deploying and serving trained machine learning pipelines on local or remote hardware.
• Model Deployment Management - Manages the operational lifecycle of models, from staging to production promotion in a registry.
• Machine Learning Pipelines - Automates the sequencing of preprocessing, model selection, and evaluation for tabular data.
• AI-Assisted ML Experiment Planning - Proposes complete experimental configurations and action plans based on a dataset and a specific analytical goal.
• Artifact Logging - Associates and persists model artifacts and files with specific execution runs in local or cloud storage.
• Model Lineage Trackers - Visualizes the provenance and relationships between experiment runs, model versions, and deployments.
• GPU Resource Scaling - Manages dedicated GPU worker pools and hardware resource limits for accelerated machine learning workloads.
• ML Management APIs - Manages models and deployments through a programmatic interface secured by tokens and API keys.
• Model Configuration - Provides a central catalog to manage model hyperparameters, search spaces, and production deployment status.
• Model Retraining Schedulers - Automates the refresh of machine learning models on a fixed timetable to prevent performance decay.
• Model Ensembling - Combines multiple models using bagging, boosting, or stacking to improve overall predictive accuracy.
• Model Performance Visualizations - Generates diagnostic plots like ROC curves and confusion matrices to evaluate model behavior.
• Model Explainability - Provides transparency by explaining model decisions and quantifying feature importance via value-based analysis.
• Model Refitting - Allows refitting a selected model on the entire dataset to maximize the information used for final production training.
• Statistical Analysis - Performs common statistical tests including t-tests and regression, presenting results via interactive visual cards.
• Drift Detection - Monitors model performance and detects data distribution shifts to trigger automatic retraining.
• Model Performance Decay Detection - Provides automated detection of model performance decay by analyzing feature distributions over time.
• Experiment Tracking - Logs and persists fitted pipelines and experiment states to enable reproduction and reuse of machine learning workflows.
• Dataset Versioning Platforms - Tracks historical versions of datasets using schema-aware versioning to ensure machine learning reproducibility.
• Time-Series Statistical Profiling - Runs normality, stationarity, and white-noise tests to validate statistical assumptions for time-series forecasting.
• Interactive Notebook Environments - Provides an integrated notebook environment for exploratory data analysis and manual model prototyping.
• Job Schedulers - Automates the re-execution of machine learning experiments on a fixed timetable using new dataset snapshots.
• Background Job Queues - Distributes asynchronous machine learning jobs across worker pools using a queued system to handle heavy compute workloads.
• Infrastructure Provisioning Tools - Automates the deployment of a full cloud stack including load balancers and clusters via declarative configuration.
• Container Deployment - Orchestrates a multi-service stack including UI, API, and workers using container composition for easy deployment.
• Control Planes - Provides a decoupled visual interface for managing system configurations and orchestrating machine learning experiments.
• Data Drift Detectors - Identifies model performance decay by comparing statistical snapshots of training data against real-time prediction logs.
• Kubernetes Deployments - Provides Helm charts and configurations for installing the platform into Kubernetes clusters with ingress and storage support.
• Web Service Deployments - Automates the deployment and monitoring of trained model pipelines as scalable services on private infrastructure.
• Pipeline Rollbacks - Tracks pipeline versions to allow reverting a production deployment to a previous stable version.
• Metric and Performance Monitors - Sets alert rules based on performance metrics and thresholds to notify users of model behavior.
• Deployment Rollbacks - Provides mechanisms to revert production model deployments to previous stable versions.
• Automated Machine Learning - Low-code library for automating end-to-end ML workflows.
• General Machine Learning - Low-code library for automating machine learning workflows.
• Machine Learning - Low-code machine learning library for rapid experimentation.
• Machine Learning Libraries - Low-code AutoML platform.
• Training and Orchestration - Low-code library for training and deploying ML models.

Star history