Automate Machine Learning Tasks

Auto ML - Automated Machine Learning Platform

An intelligent automated machine learning platform that provides comprehensive data analysis, preprocessing, model selection, and hyperparameter tuning capabilities through Model Context Protocol (MCP) tools.

🚀 Features

📊 Data Analysis & Exploration

• Data Information: Get comprehensive dataset statistics including shape, memory usage, data types, and missing values
• CSV Reading: Efficient CSV file reading with pandas and pyarrow support
• Correlation Analysis: Visualize correlation matrices for numerical and categorical variables
• Outlier Detection: Identify and visualize outliers in your datasets

🔧 Data Preprocessing

• Automated Preprocessing: Handle missing values, encode categorical variables, and scale numerical features
• Feature Engineering: Prepare features for both regression and classification problems
• Data Validation: Check for duplicates and data quality issues

🤖 Machine Learning Models

• Multiple Algorithms: Support for various ML algorithms including:
  • Regression: Linear Regression, Ridge, Lasso, ElasticNet, Random Forest, XGBoost, SVR, KNN, CatBoost
  • Classification: Logistic Regression, Ridge Classifier, Random Forest, XGBoost, SVM, KNN, Decision Tree, Naive Bayes, CatBoost

📈 Model Evaluation & Visualization

• Performance Metrics:
  • Regression: R², MAE, MSE
  • Classification: Accuracy, F1-Score
• Confusion Matrix Visualization: For classification problems
• Model Comparison: Compare multiple models side-by-side

⚙️ Hyperparameter Tuning

• Automated Tuning: Optimize model hyperparameters using advanced search algorithms
• Customizable Scoring: Choose from various evaluation metrics
• Trial Management: Control the number of optimization trials

📁 Project Structure

AutoML/
├── data/                   # Sample datasets
│   ├── Ai.csv
│   ├── Calories.csv
│   ├── Cost.csv
│   ├── Digital.csv
│   ├── Electricity.csv
│   ├── ford.csv
│   ├── Habits.csv
│   ├── heart.csv
│   ├── Lifestyle.csv
│   ├── Mobiles.csv
│   ├── Personality.csv
│   ├── Salaries.csv
│   ├── Shopper.csv
│   ├── Sleep.csv
│   ├── cat.csv
│   ├── test.csv
│   └── train.csv
├── tools/
│   └── all_tools.py       # MCP tool definitions
├── utils/
│   ├── before_model.py        # Feature preparation
│   ├── details.py             # Data information
│   ├── external_test.py       # External data test with XGBoost
│   ├── feature_importance.py  # Feature importance analysis
│   ├── hyperparameter.py      # Hyperparameter tuning
│   ├── model_selection.py     # Model selection and evaluation
│   ├── prediction.py          # Prediction utilities
│   ├── preprocessing.py       # Data preprocessing
│   ├── read_csv_file.py       # CSV reading utilities
│   └── visualize_data.py      # Visualization functions
├── main.py                # Application entry point
├── server.py              # MCP server configuration
├── requirements.txt       # Python dependencies
└── README.md             # This file

🛠️ Installation

Prerequisites

• Python 3.8 or higher
• pip or uv package manager

Setup

1. Clone the repository

   git clone https://github.com/emircansoftware/AutoML.git
   cd AutoML
   

2. Install dependencies

   # Using pip
   pip install -r requirements.txt
   pip install uv
   

Using with Claude Desktop

1. Data Path Setting

In utils/read_csv_file.py, update the path variable to match your own project directory on your computer:

# Example:
path = r"C:\\YOUR\\PROJECT\\PATH\\AutoML\\data"

2. Claude Desktop Configuration

In Claude Desktop, add the following block to your claude_desktop_config.json file and adjust the paths to match your own system:

{
  "mcpServers": {
    "AutoML": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\YOUR\\PROJECT\\PATH\\AutoML",
        "run",
        "main.py"
      ]
    }
  }
}

You can now start your project from Claude Desktop.

📋 Dependencies

• MCP Framework: mcp[cli]>=1.9.4 - Model Context Protocol for tool integration
• Data Processing: pandas>=2.3.0, pyarrow>=20.0.0, numpy>=2.3.1
• Machine Learning: scikit-learn>=1.3.0, xgboost>=2.0.0, lightgbm>=4.3.0
• Additional ML: catboost (for CatBoost models)

🎯 Usage

Starting the MCP Server

from server import mcp

# Run the server
mcp.run()

Available Tools

The platform provides the following MCP tools:

Data Analysis Tools

• information_about_data(file_name): Give detailed information about the data
• reading_csv(file_name): Read the csv file
• visualize_correlation_num(file_name): Visualize the correlation matrix for numerical columns
• visualize_correlation_cat(file_name): Visualize the correlation matrix for categorical columns
• visualize_correlation_final(file_name, target_column): Visualize the correlation matrix after preprocessing
• visualize_outliers(file_name): Visualize outliers in the data
• visualize_outliers_final(file_name, target_column): Visualize outliers after preprocessing

Preprocessing Tools

• preprocessing_data(file_name, target_column): Preprocess the data (remove outliers, fill nulls, etc.)
• prepare_data(file_name, target_column, problem_type): Prepare the data for models (encoding, scaling, etc.)

Model Training & Evaluation

• models(problem_type, file_name, target_column): Select and evaluate models based on problem type
• visualize_accuracy_matrix(file_name, target_column, problem_type): Visualize the confusion matrix for predictions
• best_model_hyperparameter(model_name, file_name, target_column, problem_type, n_trials, scoring, random_state): Tune the hyperparameters of the best model
• test_external_data(main_file_name, target_column, problem_type, test_file_name): Test external data with the best model and return predictions
• predict_value(model_name, file_name, target_column, problem_type, n_trials, scoring, random_state, input): Predict the value of the target column for new input
• feature_importance_analysis(file_name, target_column, problem_type): Analyze the feature importance of the data using XGBoost

Example Workflow

# 1. Analyze your data
info = information_about_data("data/heart.csv")

# 2. Preprocess the data
preprocessed = preprocessing_data("data/heart.csv", "target")

# 3. Prepare features for classification
features = prepare_data("data/heart.csv", "target", "classification")

# 4. Train and evaluate models
results = models("classification", "data/heart.csv", "target")

# 5. Visualize results
confusion_matrix = visualize_accuracy_matrix("data/heart.csv", "target", "classification")

# 6. Optimize best model
best_model = best_model_hyperparameter("RandomForestClassifier", "data/heart.csv", "target", "classification", 100, "accuracy", 42)

📊 Sample Datasets (All CSV datasets are from Kaggle.)

The project includes various sample datasets for testing:

• heart.csv: Heart disease prediction dataset
• Salaries.csv: Salary prediction dataset
• Calories.csv: Calorie prediction dataset
• Personality.csv: Personality analysis dataset
• Digital.csv: Digital behavior dataset
• Lifestyle.csv: Lifestyle analysis dataset
• Mobiles.csv: Mobile phone dataset
• Habits.csv: Habit analysis dataset
• Sleep.csv: Sleep pattern dataset
• Cost.csv: Cost analysis dataset
• ford.csv: Ford car dataset
• Ai.csv: AI-related dataset
• cat.csv: Cat-related dataset

🔧 Configuration

Environment Variables

• Set your preferred random seed for reproducible results
• Configure MCP server settings in server.py

Customization

• Add new ML algorithms in utils/model_selection.py
• Extend preprocessing steps in utils/preprocessing.py
• Create custom visualization functions in utils/visualize_data.py

🤝 Contributing

We welcome contributions! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

📝 License

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

🙏 Acknowledgments

• Model Context Protocol for the MCP framework
• scikit-learn for machine learning algorithms
• XGBoost for gradient boosting
• CatBoost for categorical boosting
• pandas for data manipulation

📞 Support

If you encounter any issues or have questions:

1. Check the Issues page
2. Create a new issue with detailed information
3. Contact the maintainers