mlplt 0.2.0

The complete ML toolkit — EDA, cleaning, training, explainability, deployment

mlpilot 🚀

mlpilot is the complete Python ML toolkit — what currently takes 30–40 hours of repetitive boilerplate takes 5–10 minutes. One import. Every tool you need. Full explainability.

import mlpilot as ml

eda    = ml.analyze(df, target='churn')          # 12-section EDA report
clean  = ml.clean(df, target='churn')            # auto null/outlier/dtype fixing
feats  = ml.features(clean.df, target='churn')  # leakage-safe feature pipeline
board  = ml.baseline(X_train, y_train)           # 15+ model leaderboard in 2 min
tuned  = ml.tune('lgbm', X_train, y_train)       # Bayesian hyperparameter search
exp    = ml.explain(tuned.best_model, X_train)   # SHAP global + local explanations
api    = ml.deploy(tuned.best_model)             # FastAPI + Docker in 5 minutes
api.serve(port=8000)                              # → localhost:8000/predict

Why mlpilot?

Feature	mlpilot	ydata-profiling	sweetviz	PyCaret	SHAP
Smart EDA report	✅	✅	✅	❌	❌
Auto data cleaning	✅	❌	❌	Partial	❌
Multi-model baseline	✅	❌	❌	✅	❌
Hyperparameter tuning	✅	❌	❌	✅	❌
Model explainability	✅	❌	❌	❌	✅
Time series	✅	❌	❌	✅	❌
NLP pipeline	✅	❌	❌	✅	❌
API deployment	✅	❌	❌	❌	❌
AI data analyst	✅	❌	❌	❌	❌
Undo / diff reports	✅	❌	❌	❌	❌

Installation

# Core (EDA, cleaning, validation, features, training)
pip install mlplt

# With specific extras
pip install mlplt[xgb,lgbm,shap,optuna]

# Everything
pip install mlplt[full]

Available extras: xgb, lgbm, shap, optuna, prophet, nlp, imb, deploy, ai, full

Modules

Module	Function	Description
SmartEDA	ml.analyze(df)	12-section EDA report with plots
AutoCleaner	ml.clean(df)	Auto null/outlier/dtype fixing with undo
DataValidator	ml.validate(df)	Schema, leakage, drift detection
FeatureForge	ml.features(df)	Leakage-safe encoding + scaling pipeline
BaselineBlitz	ml.baseline(X, y)	15+ model comparison leaderboard
EvalSuite	ml.evaluate(model, X, y)	All metrics + diagnostic plots
HyperX	ml.tune(model, X, y)	Bayesian hyperparameter optimization
Explainer	ml.explain(model, X)	SHAP global + local + what-if
BalanceKit	ml.balance(X, y)	Auto SMOTE/ADASYN/class_weight
TimeSense	ml.forecast(df)	Multi-model time series forecasting
TextML	ml.text_classify(df)	NLP classification + embeddings
LaunchPad	ml.deploy(model)	FastAPI + Docker generation
AIAnalyst	ml.analyst(df)	Ask questions in plain English

Quick Start — Churn Prediction

import mlpilot as ml
import pandas as pd

df = pd.read_csv('churn.csv')

# 1. Understand your data
eda = ml.analyze(df, target='Churn', report_format='html')

# 2. Clean it
df_clean = ml.clean(df, target='Churn').df

# 3. Engineer features (leakage-safe)
feats = ml.features(df_clean, target='Churn')
X_train, X_test, y_train, y_test = ml.split(feats, test_size=0.2, stratify=True)

# 4. Handle imbalance
bal = ml.balance(X_train, y_train)

# 5. Find the best model
board = ml.baseline(bal.X_resampled, bal.y_resampled, X_test=X_test, y_test=y_test)
board.leaderboard.print()

# 6. Tune + evaluate
tuned = ml.tune('lgbm', bal.X_resampled, bal.y_resampled, time_budget=300)
eval_r = ml.evaluate(tuned.best_model, X_test, y_test, optimize_threshold=True)

# 7. Explain
exp = ml.explain(tuned.best_model, X_train, X_test)
exp.feature_importance()

# 8. Deploy
ml.deploy(tuned.best_model, X_sample=X_test.iloc[:10]).serve(port=8000)

Documentation

Full API reference: mlpilot.readthedocs.io

Contributing

1. Fork the repo
2. pip install -e ".[dev]"
3. pre-commit install
4. Make your changes + add tests
5. pytest tests/ --cov=mlpilot
6. Open a pull request

License

MIT — see LICENSE.