featuremind v3.1.1

One-line AutoML with built-in reliability, leakage detection & explainability

🎬 Demo Video

Try it in 30 seconds

pip install featuremind==3.1.1

# After installation import the featuremind
import featuremind as fm
fm.analyze("data.csv")

One-line AutoML with Built-in Reliability, Leakage Detection & Explainability

---

---

What is featuremind?

featuremind is a one-line AutoML library that handles the complete machine learning pipeline — from raw CSV to production-ready model — with built-in reliability checking, leakage detection, and feature engineering.

##  Tested Datasets

featuremind v3.1 has been verified on:

| Dataset                          | Type           | Score         | Notes                                               |
| -------------------------------- | -------------- | ------------- | --------------------------------------------------- |
| Telecom Churn (7,043 rows)       | Classification | 85.7% F1      | ✅ Stable, well-balanced                             | |
| Heart Failure Medical            | Classification | ~80% Accuracy | ✅ Works                                             |
| House Prices                     | Regression     | R² reported   | ✅ Works                                             |
| Generic CSVs                     | Auto-detected  | Auto-detected | ✅ Works                                             |

---

Key Features

Auto ML Pipeline

• Loads and cleans any CSV automatically
• Detects target column, task type (classification / regression), and data issues
• Trains 6 models: LogisticRegression, RandomForest, GradientBoosting, XGBoost, LightGBM, CatBoost
• Picks best model using cross-validation
• Auto hyperparameter tuning (RandomizedSearchCV)

Leakage Guard (Core Feature)

• Detects if any feature formula references the target column
• Flags columns with suspiciously high correlation with target (>0.95)
• Smart ID detection (non-generalizable columns)
• Warns user before model training (no silent failures)

Reliability Engine

• Detects unrealistic scores (>0.98)

• Adjusts confidence level automatically:

  • 0.99 → Low confidence

  • 0.98 → Medium

• Highlights possible issues:

  • Data leakage
  • Overfitting
  • Sampling bias

Class Imbalance Handling

• Detects imbalance automatically
• Applies SMOTE (if available)
• Falls back to class weights
• Switches evaluation metric to F1 when needed

SHAP Explainability

• Computes SHAP values for model explainability
• Displays top features influencing predictions
• Helps identify real business drivers

Feature Engineering (Multi-layer)

• Domain-aware features: Telecom · Medical · Real Estate · Finance · HR
• Interactions, ratios, log transforms, polynomial features
• Only surfaces features that improve performance

Production Pipeline

• Save trained model + preprocessing pipeline
• Load and predict on new/unseen data
• Handles missing columns and unseen categories

Experiment Tracking

• Logs every run automatically
• Leaderboard of models and scores
• Export results to CSV

REST API (Optional)

• FastAPI-based prediction server
• Ready-to-use endpoints for deployment

---

Why featuremind?

Capability	featuremind	Typical AutoML Tools
One-line usage	✅	❌
Leakage detection	✅	❌
Reliability scoring	✅	❌
SHAP explainability	✅	⚠️
Production pipeline	✅	✅

---

Installation

pip install featuremind

# (Recommended) Install advanced ML libraries
pip install xgboost lightgbm catboost shap imbalanced-learn

# Optional API support
pip install fastapi uvicorn python-multipart

---

Quick Start

import featuremind as fm

fm.analyze("data.csv")
fm.check_leakage("data.csv", target="Churn")

pipeline = fm.train("data.csv", target="Churn")
pipeline.save("churn_pipeline")

pipeline = fm.load_pipeline("churn_pipeline")
results = pipeline.predict_df(new_data)

fm.get_tracker().leaderboard()

fm.serve("churn_pipeline/", port=8000)

---

Example Output

 featuremind v3.1.1 — Starting Analysis
 Best Model   : LightGBM
 Score        : 0.8569 (F1-weighted)
 Confidence   : High 
 Leakage      : None detected

---

Project Structure

featuremind_project/
│
├── featuremind/
│   ├── analyzer.py
│   ├── feature_engineer.py
│   ├── evaluator.py
│   ├── leakage_guard.py
│   ├── importance.py
│   ├── reporter.py
│   ├── html_reporter.py
│   ├── insights.py
│   ├── pipeline.py
│   ├── tracker.py
│   └── api.py
│
├── setup.py
├── requirements.txt
├── test.py
└── README.md

---

Notes

• High accuracy (>0.98) may indicate:

  • Data leakage
  • Highly separable datasets
  • Sampling bias

• Always validate models on unseen data.

---

Output Files

• featuremind_report.html → Full analysis report
• featuremind_report.png → Feature visualization
• enhanced_data.csv → Dataset with engineered features
• featuremind_experiments.csv → Experiment logs
• pipeline/ → Saved production model

---

Use Cases

• Telecom churn prediction
• Fraud detection
• Healthcare predictions
• Real estate pricing
• HR analytics
• Any tabular ML problem

---

Why Developers Love featuremind

• Go from raw data → model in 1 line
• Built-in leakage detection (rare in AutoML)
• Explainable AI (SHAP) included by default
• Reliability scoring (not just accuracy)
• Direct production pipeline export

Not just AutoML — this is AutoML + Trust Layer

---

Roadmap

• Time-series support
• Deep learning integration
• Streamlit dashboard
• Cloud deployment

---

License

MIT License

---

Author

Niveditha — Data Scientist & ML Engineer

LinkedIn: https://www.linkedin.com/in/niveditha-89ba04356/
PyPI: https://pypi.org/project/featuremind/
GitHub: https://github.com/Nivedithagowda2/featuremind

---

⭐ If this project helps you, consider giving it a star!