mkyz 0.2.3

MKYZ is a Python library for ML and data science tasks.

MKYZ - Machine Learning Library

MKYZ is a comprehensive Python machine learning library designed to simplify data processing, model training, evaluation, and visualization tasks. Built on top of scikit-learn, it provides a unified API for common ML workflows.

📚 Examples

You can find comprehensive Jupyter notebooks in the examples/ directory:

• 01_quickstart.ipynb: End-to-end basic ML workflow.
• 02_data_profiling_and_eda.ipynb: Data profiling and HTML reports.
• 03_feature_engineering.ipynb: Advanced feature creation.
• 04_automl_and_optimization.ipynb: AutoML and hyperparameter tuning.
• 05_model_persistence_and_reports.ipynb: Saving models and rich reports.

---

🏗️ Architecture

Core Capabilities

• 🔄 Data Preparation - Automatic handling of missing values, outliers, and categorical encoding
• 🎯 Model Training - Support for 20+ classification, regression, and clustering algorithms
• 📊 AutoML - Automatic model selection and hyperparameter optimization
• 📈 Evaluation - Comprehensive metrics with 10 cross-validation strategies
• 🎨 Visualization - 40+ built-in plotting functions for EDA and model results

New in v0.2.0

• 💾 Model Persistence - Save and load models with metadata
• 🔧 Feature Engineering - Polynomial, datetime, lag, and rolling features
• 📝 Auto Reports - Generate HTML reports with one line of code
• ⚡ Parallel Processing - Built-in utilities for faster training
• 🛡️ Robust Error Handling - Custom exceptions for better debugging

📦 Installation

pip install mkyz

From Source

git clone https://github.com/mmustafakapici/mkyz.git
cd mkyz
pip install -e .

Dependencies

pandas, scikit-learn, numpy, matplotlib, seaborn, 
plotly, xgboost, lightgbm, catboost, rich, mlxtend

🚀 Quick Start

Basic Usage (Original API)

import mkyz

# 1. Prepare data
data = mkyz.prepare_data('dataset.csv', target_column='price')

# 2. Train model
model = mkyz.train(data, task='classification', model='rf')

# 3. Make predictions
predictions = mkyz.predict(data, model)

# 4. Evaluate
scores = mkyz.evaluate(data, predictions)
print(scores)

# 5. Visualize
mkyz.visualize(data)

AutoML - Find the Best Model

import mkyz

data = mkyz.prepare_data('dataset.csv', target_column='target')

# Automatically train and compare all models
best_model = mkyz.auto_train(
    data, 
    task='classification',
    optimize_models=True,
    optimization_method='bayesian'
)

New Modular API (v0.2.0)

import mkyz

# Configure globally
mkyz.set_config(random_state=42, n_jobs=-1, verbose=1)

# Load data flexibly
df = mkyz.load_data('data.csv')  # Also supports Excel, JSON, Parquet

# Validate dataset
validation = mkyz.validate_dataset(df, target_column='target')
if not validation['is_valid']:
    print(validation['issues'])

# Feature Engineering
fe = mkyz.FeatureEngineer()
df = fe.create_datetime_features(df, 'date_column')
df = fe.create_polynomial_features(df, ['age', 'income'], degree=2)

# Select best features
selected = mkyz.select_features(X, y, k=10, method='mutual_info')

# Advanced Cross-Validation
results = mkyz.cross_validate(
    model, X, y,
    cv=mkyz.CVStrategy.STRATIFIED,
    n_splits=5,
    return_train_score=True
)
print(f"Mean accuracy: {results['mean_test_score']:.4f}")

# Save trained model
mkyz.save_model(model, 'models/my_model', metadata={'version': '1.0'})

# Load model later
model = mkyz.load_model('models/my_model.joblib')

# Generate comprehensive report
report = mkyz.ModelReport(model, X_test, y_test, task='classification')
report.generate()
report.export_html('reports/model_report.html')
print(report.summary())

📚 Documentation

Modules Overview

Module	Description
mkyz.core	Configuration, exceptions, base classes
mkyz.data	Data loading, preprocessing, feature engineering
mkyz.evaluation	Metrics, cross-validation, reporting
mkyz.persistence	Model saving and loading
mkyz.utils	Logging and parallel processing utilities

Detailed Guides

• Installation Guide
• Quick Start Tutorial
• Data Preparation Guide
• Feature Engineering Guide
• Model Training Guide
• API Reference

🔧 Supported Models

Classification

Model	Key	Description
Random Forest	rf	Ensemble of decision trees
Logistic Regression	lr	Linear classification
SVM	svm	Support Vector Machine
KNN	knn	K-Nearest Neighbors
Decision Tree	dt	Single decision tree
Naive Bayes	nb	Probabilistic classifier
Gradient Boosting	gb	Boosted trees
XGBoost	xgb	Extreme Gradient Boosting
LightGBM	lgbm	Light Gradient Boosting
CatBoost	catboost	Categorical Boosting

Regression

Model	Key	Description
Random Forest	rf	Ensemble regressor
Linear Regression	lr	OLS regression
SVR	svm	Support Vector Regression
KNN	knn	K-Nearest Neighbors
Decision Tree	dt	Single decision tree

Clustering

Model	Key	Description
K-Means	kmeans	Centroid-based
DBSCAN	dbscan	Density-based
Agglomerative	agglomerative	Hierarchical
GMM	gmm	Gaussian Mixture
Mean Shift	mean_shift	Mode-seeking

Dimensionality Reduction

Model	Key	Description
PCA	pca	Principal Component Analysis
SVD	svd	Truncated SVD
NMF	nmf	Non-negative Matrix Factorization

📊 Cross-Validation Strategies

from mkyz import cross_validate, CVStrategy

# Available strategies
strategies = [
    CVStrategy.KFOLD,              # Standard K-Fold
    CVStrategy.STRATIFIED,         # Stratified K-Fold (default)
    CVStrategy.TIME_SERIES,        # Time Series Split
    CVStrategy.GROUP,              # Group K-Fold
    CVStrategy.REPEATED,           # Repeated K-Fold
    CVStrategy.REPEATED_STRATIFIED,# Repeated Stratified
    CVStrategy.LEAVE_ONE_OUT,      # Leave-One-Out
    CVStrategy.SHUFFLE,            # Shuffle Split
    CVStrategy.STRATIFIED_SHUFFLE  # Stratified Shuffle
]

# Usage
results = cross_validate(model, X, y, cv=CVStrategy.TIME_SERIES, n_splits=5)

🔧 Configuration

import mkyz

# View current config
print(mkyz.get_config().to_dict())

# Update config
mkyz.set_config(
    random_state=42,
    n_jobs=-1,
    cv_folds=5,
    verbose=1,
    dark_mode=True
)

Available Settings

Setting	Default	Description
random_state	42	Random seed for reproducibility
n_jobs	-1	Parallel jobs (-1 = all CPUs)
cv_folds	5	Default CV folds
test_size	0.2	Train/test split ratio
verbose	1	Verbosity level
optimization_method	'grid_search'	'grid_search' or 'bayesian'
missing_value_strategy	'mean'	'mean', 'median', 'mode', 'drop'
outlier_strategy	'remove'	'remove', 'cap', 'keep'

🛡️ Error Handling

from mkyz import (
    MKYZError,           # Base exception
    DataValidationError, # Data issues
    ModelNotTrainedError,# Model not fitted
    UnsupportedTaskError,# Invalid task type
    PersistenceError     # Save/load failures
)

try:
    model = mkyz.load_model('nonexistent.joblib')
except PersistenceError as e:
    print(f"Failed to load model: {e}")

📈 Visualization

import mkyz

# EDA visualizations
mkyz.visualize(data, plot_type='histogram')
mkyz.visualize(data, plot_type='correlation')
mkyz.visualize(data, plot_type='boxplot')

# Available plot types:
# histogram, bar, box, violin, pie, scatter, line,
# heatmap, pair, swarm, strip, kde, ridge, density,
# joint, regression, residual, qq, ecdf, dendrogram...

🤝 Contributing

Contributions are welcome! Please read our Contributing Guide.

# Clone the repository
git clone https://github.com/mmustafakapici/mkyz.git

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest tests/

📄 License

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

👤 Author

Mustafa Kapıcı

• Email: [e-mail]
• GitHub: @mmustafakapici

🙏 Acknowledgments

• Built on top of scikit-learn
• Boosting models from XGBoost, LightGBM, CatBoost
• Visualization powered by Matplotlib, Seaborn, Plotly

---

Made with ❤️ in Turkey