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

Installation

From Source Code

git clone https://github.com/emircansoftware/AutoML.git
cd AutoML
pip install -r requirements.txt
pip install uv

Configuration

Claude Desktop

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

Available Tools

Tool	Description
information_about_data	Provides detailed information about the data
reading_csv	Reads a CSV file
visualize_correlation_num	Visualizes correlation matrix for numerical columns
visualize_correlation_cat	Visualizes correlation matrix for categorical columns
visualize_correlation_final	Visualizes correlation matrix after preprocessing
visualize_outliers	Visualizes outliers in the data
visualize_outliers_final	Visualizes outliers after preprocessing
preprocessing_data	Preprocesses data (removing outliers, filling missing values, etc.)
prepare_data	Prepares data for models (encoding, scaling, etc.)
models	Selects and evaluates models based on task type
visualize_accuracy_matrix	Visualizes confusion matrix for predictions
best_model_hyperparameter	Tunes hyperparameters of the best model
test_external_data	Tests external data with the best model and returns predictions
predict_value	Predicts target column value for new input data
feature_importance_analysis	Analyzes feature importance in the data using XGBoost

Features

• Comprehensive dataset statistics including size, memory usage, data types, and missing values
• Efficient CSV file reading with pandas and pyarrow support
• Correlation analysis and visualization for numerical and categorical variables
• Outlier detection and visualization
• Automated preprocessing with missing value handling, categorical feature encoding, and scaling
• Support for multiple machine learning algorithms including Linear Regression, Ridge, Lasso, ElasticNet, Random Forest, XGBoost, SVR, KNN, CatBoost
• Classification algorithms including Logistic Regression, Ridge Classifier, Random Forest, XGBoost, SVM, KNN, Decision Tree, Naive Bayes, CatBoost
• Performance metrics for regression (R², MAE, MSE) and classification (Accuracy, F1-Score)
• Confusion matrix visualization for classification tasks
• Model comparison capabilities

Usage Examples

Analyze dataset statistics and missing values for heart.csv

Preprocess data by handling missing values and outliers for target column

Train and compare multiple classification models on heart disease dataset

Visualize correlation matrix for numerical features in the dataset

Optimize hyperparameters for RandomForestClassifier with 100 trials

Notes

Requires Python 3.8+. You must update the data path in utils/read_csv_file.py according to your project directory. Includes 16 sample datasets from Kaggle for testing. The server should be configured with correct local paths in Claude Desktop configuration.