MLTrackFlow 🚀

A User-Friendly Python Library for Making Machine Learning Training Processes Transparent and Traceable

MLTrackFlow is a beginner-friendly Python package that enables you to track, record, and visualize your ML model development process step by step.

---

🌟 Why MLTrackFlow?

🎯 Get Started in One Line

from mltrackflow import ExperimentTracker

tracker = ExperimentTracker(experiment_name="my_project")
with tracker.start_run("first_experiment"):
    tracker.log_model_metrics(model, X_test, y_test)  # Automatic!

✨ Key Features

• 🎓 Perfect for Beginners: Simple API, automatic logging, plenty of examples
• 📊 Automatic Metric Tracking: Accuracy, precision, recall, F1 calculated automatically
• 🔄 Pipeline Management: Organize all steps from data preparation to model
• 📈 Rich Visualization: Confusion matrix, learning curves, feature importance
• 🏆 Model Comparison: Easily compare different models
• 📄 HTML Reports: Professional reports with one click
• 💾 Model Versioning: Organize all your models
• 🔒 Data Tracking: Track changes with automatic data hashing

🚀 Quick Start

Installation

pip install mltrackflow

Your First Experiment (60 seconds!)

from mltrackflow import ExperimentTracker
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier

# Prepare data
data = load_iris()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

# Start tracker
tracker = ExperimentTracker(experiment_name="iris_demo")

# Train and log
with tracker.start_run("random_forest"):
    # Log parameters
    tracker.log_params({"n_estimators": 100, "max_depth": 5})
    
    # Train model
    model = RandomForestClassifier(n_estimators=100, max_depth=5)
    model.fit(X_train, y_train)
    
    # Automatically calculate and log metrics
    tracker.log_model_metrics(model, X_test, y_test)
    
    # Save model
    tracker.save_model(model, "my_model")

# Generate HTML report
tracker.generate_report()
print("✅ Report created: experiments/iris_demo/experiment_report.html")

Output:

🚀 Run started: random_forest
📊 Metric logged: accuracy = 0.9667
📊 Metric logged: precision = 0.9722
📊 Metric logged: recall = 0.9667
📊 Metric logged: f1_score = 0.9667
✅ Run completed: random_forest

📚 Feature Details

1️⃣ Modular Workflow with Pipeline

from mltrackflow import MLPipeline, PipelineStep
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA

# Create pipeline
pipeline = MLPipeline(name="data_pipeline", tracker=tracker)

# Add steps
pipeline.add_step(PipelineStep(name="scaler", transformer=StandardScaler()))
pipeline.add_step(PipelineStep(name="pca", transformer=PCA(n_components=2)))
pipeline.add_step(PipelineStep(name="model", model=RandomForestClassifier()))

# Train
with tracker.start_run("pipeline_demo"):
    pipeline.fit(X_train, y_train)
    predictions = pipeline.predict(X_test)

# Visualize
pipeline.visualize_steps(output_path="pipeline.png")

2️⃣ Model Comparison

from mltrackflow import ModelComparator

# Try different models
models = {
    "rf": RandomForestClassifier(n_estimators=100),
    "svm": SVC(kernel='rbf'),
    "logistic": LogisticRegression(),
}

for name, model in models.items():
    with tracker.start_run(name):
        model.fit(X_train, y_train)
        tracker.log_model_metrics(model, X_test, y_test)

# Compare
comparator = ModelComparator(tracker=tracker)
comparator.compare_runs()
comparator.print_comparison_table()

# Find best
best = comparator.get_best_model(metric="accuracy", maximize=True)
print(f"🏆 Best model: {best}")

3️⃣ Visualization

from mltrackflow import Visualizer

viz = Visualizer(tracker=tracker)

# Confusion matrix
viz.plot_confusion_matrix(y_test, predictions)

# Feature importance
viz.plot_feature_importance(model, feature_names)

# Model comparison
viz.plot_metrics_comparison(
    run_names=["rf", "svm", "logistic"],
    metrics=["accuracy", "f1_score"]
)

🆚 Comparison with Other Tools

Feature	MLflow	W&B	MLTrackFlow
Installation	Complex	Registration required	pip install ✅
Learning Curve	Medium	Medium	Easy 🎓
Local Execution	✅	Limited	✅
Pipeline Support	❌	❌	✅
Auto Metrics	Limited	Limited	Fully Automatic 🤖
Beginner Friendly	⚠️	⚠️	✅

💡 Command Line Usage

# Start new experiment
mltrackflow init --name my_experiment

# List experiments
mltrackflow list

# Generate report
mltrackflow report --experiment iris_demo

# Compare models
mltrackflow compare --experiment iris_demo --runs rf svm logistic

📖 Documentation

• Quick Start Guide
• API Reference
• Example Projects

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md.

📄 License

This project is licensed under the MIT License - see LICENSE file for details.

---

🇹🇷 Türkçe Açıklama

Makine Öğrenimi Eğitim Süreçlerini Şeffaf ve İzlenebilir Hale Getiren Kullanıcı Dostu Python Kütüphanesi

MLTrackFlow, ML model geliştirme sürecinizi adım adım izlemenize, kayıt altına almanıza ve görselleştirmenize olanak tanıyan yeni başlayanlar için ideal bir Python paketidir.

🌟 Neden MLTrackFlow?

• 🎓 Yeni Başlayanlar İçin: Basit API, otomatik loglama
• 📊 Otomatik Metrik Takibi: Tüm metrikler otomatik hesaplanır
• 🔄 Pipeline Yönetimi: Veri hazırlıktan modele kadar tüm adımları organize edin
• 📈 Zengin Görselleştirme: Karmaşıklık matrisi, öğrenme eğrileri
• 🏆 Model Karşılaştırma: Farklı modelleri kolayca kıyaslayın
• 📄 HTML Raporları: Tek tıkla profesyonel raporlar

🚀 Kurulum ve Kullanım

pip install mltrackflow

from mltrackflow import ExperimentTracker
from sklearn.ensemble import RandomForestClassifier

# Tracker başlat
tracker = ExperimentTracker(experiment_name="proje_adi")

# Model eğit ve kaydet
with tracker.start_run("deneme_1"):
    model = RandomForestClassifier()
    model.fit(X_train, y_train)
    
    # Metrikleri otomatik kaydet
    tracker.log_model_metrics(model, X_test, y_test)
    
    # Modeli kaydet
    tracker.save_model(model, "modelim")

# HTML rapor oluştur
tracker.generate_report()

📚 Özellikler

Pipeline ile Modüler İş Akışı

from mltrackflow import MLPipeline, PipelineStep
from sklearn.preprocessing import StandardScaler

pipeline = MLPipeline(name="veri_pipeline")
pipeline.add_step(PipelineStep(name="olcekleme", transformer=StandardScaler()))
pipeline.add_step(PipelineStep(name="model", model=RandomForestClassifier()))

pipeline.fit(X_train, y_train)
pipeline.visualize_steps()  # Pipeline'ı görselleştir

Model Karşılaştırma

from mltrackflow import ModelComparator

# Farklı modelleri dene
for model_name, model in models.items():
    with tracker.start_run(model_name):
        model.fit(X_train, y_train)
        tracker.log_model_metrics(model, X_test, y_test)

# Karşılaştır
comparator = ModelComparator(tracker=tracker)
comparator.compare_runs()
best = comparator.get_best_model(metric="accuracy")

Görselleştirme

from mltrackflow import Visualizer

viz = Visualizer(tracker=tracker)
viz.plot_confusion_matrix(y_test, predictions)
viz.plot_feature_importance(model, feature_names)
viz.plot_metrics_comparison(["model1", "model2"])

💡 Komut Satırı

# Yeni deney başlat
mltrackflow init --name proje_adi

# Deneyleri listele
mltrackflow list

# Rapor oluştur
mltrackflow report --experiment proje_adi

🎯 Ne Görürsünüz?

Her deneyde:

• ✅ Otomatik parametre ve metrik kaydı
• ✅ Zaman damgası
• ✅ Karşılaştırma tablosu
• ✅ HTML rapor (grafiklerle)
• ✅ En iyi model otomatik seçimi

📖 Dokümantasyon

• Hızlı Başlangıç Rehberi (Türkçe)
• Örnek Projeler
• API Dökümanı

🤝 Katkıda Bulunma

Katkılarınızı bekliyoruz! CONTRIBUTING.md dosyasına bakın.

📄 Lisans

Bu proje MIT lisansı altında lisanslanmıştır.

---

Quick Links: GitHub • PyPI • Examples • Issues

Don't forget to star! ⭐ / Yıldız vermeyi unutmayın! ⭐