AutoEDA: Automated Exploratory Data Analysis & AutoML Suite

Overview

AutoEDA is a professional, user-friendly Streamlit application for automated data exploration, preprocessing, and machine learning (AutoML). It is designed for data scientists, analysts, and ML practitioners who want to:

• Quickly explore and visualize datasets
• Effortlessly preprocess data
• Automatically train, evaluate, and compare ML models with FLAML
• Interpret results with explainability and error analysis tools

---

Features

🚀 Automated EDA

• Dataset summary, missing values, data types
• Interactive visualizations (scatter, pairplot, heatmap, etc.)
• Categorical & numerical feature exploration

🛠️ Data Preprocessing

• Remove or fill missing values (mean, median, mode)
• Scale features (Standard, MinMax)
• Outlier detection (z-score)
• Column selection/removal

🤖 AutoML (FLAML)

• Supports classification & regression
• Rich model coverage: LightGBM, XGBoost, Random Forest, Extra Trees, HistGB, SVC, KNeighbors, L1/L2-regularized models, CatBoost, etc.
• Smart target/feature selection
• Cross-validation, hyperparameter tuning
• Model leaderboard
• Download predictions as CSV
• Download trained AutoML object (pickle)

📊 Results & Explainability

• Performance metrics (Accuracy, F1, MAE, R2, etc.) in a clear table
• Confusion matrix, ROC curve, regression plots
• SHAP explainability with summary plots
• Feature importance
• Error analysis: top misclassified samples/largest errors

🖥️ Professional UI/UX

• Modern Streamlit interface
• Sidebar navigation
• Tooltips, badges, and user feedback
• Robust error handling

---

Folder Structure

Auto_EDA/
├── .Streamlit/                 # Streamlit config (optional)
├── __pycache__/                # Python cache files (auto-generated)
├── automl_functions.py         # All AutoML, evaluation, explainability logic
├── data_analysis_functions.py  # EDA and visualization helpers
├── data_preprocessing_function.py # Data cleaning, scaling, outlier, etc.
├── example_dataset/
│   └── titanic.csv             # Example dataset for demo/testing
├── home_page.py                # Home tab and landing page UI
├── logs.log                    # App logs
├── main.py                     # Main Streamlit app (UI, navigation, orchestration)
├── requirements.txt            # All Python dependencies
├── test_automl.py              # Test script for AutoML module
├── README.md                   # (You are here)
├── venv/                       # Python virtual environment (optional)
└── .gitignore                  # Git ignore rules

---

Getting Started

1. Clone the repository

git clone <your-repo-url>
cd Auto_EDA

2. Install dependencies

pip install -r requirements.txt

3. Run the app

streamlit run main.py

4. Upload your dataset or use the example Titanic dataset

---

Usage Guide

• Home: Overview and quick start
• Data Exploration: Visualize and understand your data
• Data Preprocessing: Clean, scale, and prepare your data
• AutoML: Select problem type, target, features, run AutoML, and review results

---

Requirements

• Python 3.8–3.11 (FLAML may not support 3.12+)
• See requirements.txt for all packages

---

Notes

• SHAP explainability works best with tree-based models (LightGBM, XGBoost, RF, etc.)
• For reproducibility, set random seeds in your data and model configs
• Downloaded models are pickled FLAML AutoML objects

---

Contributing

Pull requests and issues are welcome! Please open an issue for bugs or feature requests.

---

Acknowledgements

• FLAML
• Streamlit
• scikit-learn
• SHAP
• Plotly