AutoML Application with Multiple Fine-Tuning Methods

Overview

This AutoML application, developed using Streamlit, provides an intuitive interface for users to perform various machine learning tasks, including classification and regression. The tool supports a wide range of machine learning models and fine-tuning methods, including manual hyperparameter selection, grid search, randomized search, and Optuna-based hyperparameter optimization. The app is deployed here : https://automl-vlquobsgcpynf3entqfvm8.streamlit.app/

Features

• File Upload Support:

  • Supports CSV and Excel (XLSX) file formats.

• Exploratory Data Analysis (EDA):

  • Display dataset shape, columns, and basic statistics.
  • Generate correlation heatmaps.

• Data Preprocessing:

  • Handle missing values using different strategies (Mean, Median, Interpolation, or Drop).
  • Encode categorical variables for both classification and regression tasks.
  • Standardize features using StandardScaler.

• Model Selection:

  • Supports a wide variety of models for classification and regression.
  • Classification models:
    • Logistic Regression
    • Decision Tree Classifier
    • Random Forest Classifier
    • SVM Classifier
    • MLP Classifier
    • KNeighbors Classifier
    • Gaussian Naive Bayes
    • AdaBoost Classifier
    • GradientBoosting Classifier
    • ExtraTrees Classifier
  • Regression models:
    • Linear Regression
    • Ridge Regression
    • Lasso Regression
    • ElasticNet Regression
    • Decision Tree Regressor
    • Random Forest Regressor
    • SVM Regressor
    • MLP Regressor
    • KNeighbors Regressor
    • AdaBoost Regressor
    • GradientBoosting Regressor
    • ExtraTrees Regressor

• Hyperparameter Tuning:

  • Manual hyperparameter tuning via sliders and dropdowns.
  • Automated fine-tuning using:
    • Grid Search
    • Randomized Search
    • Optuna Optimization

• Model Training and Testing:

  • Train models with user-defined hyperparameters or automated fine-tuning.
  • Test trained models and display performance metrics.
    • Classification metrics: Accuracy, Precision, Recall, F1-Score.
    • Regression metrics: Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE).

• Model Comparison:

  • Compare multiple trained models based on selected metrics.
  • Visualize model comparison results using bar charts.

• Model Download:

  • Download trained models in pickle format for future use.

Project Structure

automl_app/
├── __init__.py
├── app.py         # Main Streamlit app
├── utils/
│   ├── __init__.py
│   ├── data_handler.py  # Data loading and preprocessing
│   ├── eda.py         # Exploratory data analysis
│   ├── model_trainer.py # Model training functions
│   └── model_evaluation.py # Model evaluation and comparison

How to Use

1. Upload Dataset:
   • Select the file type (CSV or Excel) and upload your dataset.
2. Select Preferences:
   • Choose whether to enable EDA and descriptive statistics.
   • Select the task type (Classification or Regression).
   • Choose the missing value handling strategy.
3. Select Model and Hyperparameter Tuning:
   • Choose a model from the provided list.
   • Select the hyperparameter tuning approach:
     • No Tuning (Default hyperparameters)
     • Manual Hyperparams (Set values manually)
     • Complete Fine-Tuning (Grid Search, Randomized Search, or Optuna)
4. Train the Model:
   • Click the "Train Model" button to start the training process.
5. Test the Model:
   • Once training is complete, click the "Test Model" button to evaluate the model.
6. Compare Models:
   • Select trained models for comparison and visualize the results.
7. Download Trained Models:
   • Download individual models or the best-performing model in pickle format.

Requirements

• Python 3.7+
• Required Python libraries:

  pip install streamlit pandas numpy scikit-learn optuna matplotlib seaborn

Installation and Execution

1. Clone the repository or download the script.
2. Install the required libraries using the provided command.
3. Run the application using Streamlit:

   python -m streamlit run app.py

4. Access the application via the provided localhost link in your web browser.

Example Usage

1. Upload a CSV file containing a dataset.
2. Select "Classification" as the task type.
3. Choose a classification model (e.g., Random Forest Classifier).
4. Enable EDA and choose "Mean" as the missing value handling strategy.
5. Select "Complete Fine-Tuning" and choose "Grid Search" for hyperparameter optimization.
6. Train the model and download the best-performing model.

Additional Information

• This application uses Optuna for advanced hyperparameter optimization.
• Fine-tuned models are saved in session and can be downloaded individually.
• The comparison feature allows users to identify the best-performing model based on selected metrics.

Limitations

• The current version supports only CSV and Excel file formats.
• The number of trials for Optuna-based optimization is set to 10 by default.
• The application requires the dataset to be clean and properly formatted.

License

This project is licensed under the MIT License.

Acknowledgments

• The application leverages popular machine learning libraries such as scikit-learn and Optuna.
• Special thanks to the Streamlit community for providing an excellent framework for building data applications.

---

For any issues or feature requests, please contact hafizhamzakhan1997@gmail.com.