tabulaml 0.1.2

TabuLaML - Fast, explainable, no-code AutoML for tabular data

TabuLaML

Fast, explainable, no-code AutoML for tabular data.

TabuLaML transforms raw CSV data into trained, deployable machine learning models with zero configuration. Just point it at your data and go.

Installation

pip install tabulaml

With XGBoost and LightGBM support:

pip install tabulaml[all]

Quick Start

tabulaml train data.csv --target target_column

Or use in Python:

from automl import AutoMLPipeline

pipeline = AutoMLPipeline()
results = pipeline.run("data.csv", target_column="target")

What It Does

• Automatic Data Profiling - Detects column types, missing values, cardinality, and class imbalance
• Task Inference - Automatically determines if your problem is binary classification, multiclass classification, or regression
• Smart Preprocessing - Handles missing values, scales numerical features, encodes categorical variables
• Class Balancing - Addresses imbalanced datasets using SMOTE, ADASYN, and other resampling techniques
• Multi-Model Training - Trains and compares Logistic Regression, Random Forest, Gradient Boosting, XGBoost, and LightGBM
• Cross-Validation - Selects the best model using rigorous cross-validation scoring
• Export Ready - Outputs trained models (.pkl), detailed reports (HTML/JSON), confusion matrices, and feature importance charts

Output

After training, you'll find in the outputs/ folder:

outputs/
  models/      - Trained model (.pkl)
  reports/     - HTML and JSON reports
  artifacts/   - Confusion matrix, feature importance charts

Use Cases

Context	What It Optimizes
Fraud Detection	Minimize missed fraud (high recall)
Churn Prediction	Catch at-risk customers before they leave
Medical Diagnosis	Reduce false negatives for critical conditions
Spam Detection	Minimize false positives to protect legitimate emails
Loan Approval	Balance risk assessment with approval rates
Marketing Campaigns	Target likely responders efficiently
Quality Control	Catch defective products before shipping

Key Features

• Zero Configuration - Point it at a CSV and go
• Progress Bar - Clean, user-friendly progress display
• Explainable Results - Understand why the model makes predictions
• Reproducible - Set random seeds for consistent results
• Production Ready - Export models ready for deployment

Requirements

• Python 3.9+
• Works on Windows, macOS, and Linux

License

MIT License - Copyright (c) 2025 Sekpey Herbert