AutoML

A powerful web application that combines automated machine learning capabilities with interactive data analysis and natural language processing. Built with Streamlit and PyCaret, this application makes machine learning accessible to everyone.

Features

1. Data Management

• Upload your own CSV datasets with automatic encoding detection
• Support for 30+ different file encodings including UTF-8, Latin1, and various Asian encodings
• Choose from pre-loaded sample datasets
• Automatic data persistence between sessions

2. Exploratory Data Analysis

• Interactive data profiling with ydata-profiling
• Full-page report display with comprehensive statistics
• Visual data exploration
• Automatic data type detection and analysis

3. Automated Machine Learning

• Classification
  • Multiple model comparison
  • Model performance metrics
  • Model download capability
• Regression
  • Multiple model comparison
  • Model performance metrics
  • Model download capability
• Clustering
  • K-means clustering
  • Cluster assignment visualization
• Anomaly Detection
  • Isolation Forest implementation
  • Anomaly scoring
• Time Series Forecasting
  • ARIMA modeling
  • Time series visualization
  • Forecast predictions

4. Advanced Preprocessing

• TF-IDF text preprocessing with customizable features
• PCA dimensionality reduction
• Automatic feature engineering
• Support for mixed data types
• Text data transformation

5. Natural Language Interface

• Powered by Groq's Llama 3.3 70B model
• Interactive data exploration through natural language
• Intelligent data analysis responses
• Context-aware question answering

Installation

1. Clone the repository:

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

2. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Create a .streamlit/secrets.toml file with your Groq API key:

[groq]
api_key = "your-groq-api-key"

Usage

1. Start the Streamlit app:

streamlit run app.py

2. Navigate through the different sections using the sidebar:
   • Upload Dataset: Import your own CSV files with automatic encoding detection
   • Use Sample Dataset: Choose from pre-loaded datasets
   • Exploratory Data Analysis: View detailed data profiles
   • Machine Learning: Run automated ML tasks
   • Ask Questions: Interact with your data using natural language

Supported Sample Datasets

Classification

• Iris: Classic classification dataset
• Diabetes: Medical classification dataset
• Bank: Banking customer classification

Regression

• Boston: Housing price prediction
• Insurance: Insurance cost prediction

Clustering

• Wine: Wine quality clustering

Anomaly Detection

• Credit Card: Credit card fraud detection

Time Series

• Airline: Airline passenger forecasting

Model Features

Classification & Regression

• Multiple model comparison
• Performance metrics visualization
• Model download capability
• Feature importance analysis
• Optional preprocessing with TF-IDF or PCA

Clustering

• K-means implementation
• Cluster visualization
• Cluster assignment export

Anomaly Detection

• Isolation Forest algorithm
• Anomaly score calculation
• Threshold-based detection

Time Series

• ARIMA modeling
• Forecast visualization
• Time series decomposition

Requirements

The project uses the following main packages:

• streamlit
• pycaret
• ydata-profiling
• langchain
• groq
• pandas
• numpy
• scikit-learn
• plotly
• matplotlib
• seaborn

For a complete list of dependencies, see requirements.txt.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. When contributing:

1. Fork the repository
2. Create a new branch for your feature
3. Submit a pull request with a clear description of your changes

License

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