AutoML Optuna Demo

An easy and lightweight demonstration of Automated Machine Learning (AutoML) using Optuna and Scikit-learn.
This project automatically tunes hyperparameters of a RandomForestClassifier on the Iris dataset to find the best configuration for maximum accuracy.

---

📘 Overview

Manual hyperparameter tuning can be slow and confusing — especially when you don’t know which values work best for your model.

This package automates that process using Optuna, a fast and intelligent hyperparameter optimization framework.
It uses Scikit-learn’s RandomForestClassifier and Iris dataset as an example to demonstrate how AutoML works in practice.

---

🧠 What It Does

• Loads the Iris dataset using Scikit-learn
• Splits the dataset into training and testing sets
• Defines an objective function for Optuna that:
  • Chooses random values for hyperparameters (n_estimators, max_depth)
  • Trains a Random Forest model
  • Evaluates accuracy on the test set
• Optuna runs multiple trials and automatically finds the best hyperparameters
• Finally, the function returns:
  • ✅ Best hyperparameters
  • 📈 Best accuracy score

---

🧩 Installation

You can install this package directly from PyPI once uploaded:

pip install automl-optuna-demo

Or if you have cloned this repository locally, navigate to the folder and install in development mode:

pip install -e .

This ensures that you can make edits and re-run the package easily.

---

🖥️ Usage

After installation, you can use this package in two ways:

▶️ Option 1: Run directly in Python

from automl_optuna_demo.main import run_automl

# Run with default 10 trials
best_params, best_value = run_automl()

print("Best Parameters:", best_params)
print("Best Accuracy:", best_value)

▶️ Option 2: Run with more trials

You can specify how many optimization trials you want Optuna to perform (more trials = better results, but slower):

best_params, best_value = run_automl(trials=30)
print(best_params, best_value)

▶️ Option 3: Run from Command Line (if added later)

If you add a CLI entry point, you could run:

python -m automl_optuna_demo.main

---

⚙️ How It Works (Step-by-Step)

1. Dataset Loading
   Uses load_iris() from Scikit-learn, a built-in dataset with 150 samples of 3 flower species.

2. Data Splitting
   Splits data into training (80%) and testing (20%) using train_test_split().

3. Objective Function
   Inside the code, the objective function defines two tunable parameters:

   n_estimators = trial.suggest_int('n_estimators', 10, 200)
   max_depth = trial.suggest_int('max_depth', 2, 32)
   

   Optuna will choose random combinations of these values across trials.

4. Model Training
   A RandomForestClassifier is trained on the training set using these values.

5. Evaluation
   The model predicts test data and returns accuracy as a numeric score to Optuna.

6. Optimization
   Optuna runs multiple trials, compares accuracies, and selects the hyperparameters that gave the highest accuracy.

---

🧾 Example Output

Running:

from automl_optuna_demo.main import run_automl
params, score = run_automl(trials=20)
print(params)
print(score)

might produce:

{'n_estimators': 132, 'max_depth': 7}
0.9666666667

Meaning Optuna found the best configuration after 20 trials that gave about 96.7% accuracy.

---

📂 Project Structure

automl_optuna_demo/
├── __init__.py          # Makes it a Python package
├── main.py              # Core logic for AutoML optimization
setup.py                 # Packaging configuration for PyPI
setup.cfg                # Additional package metadata
README.md                # You are here

---

🧰 Requirements

This project depends on:

• Optuna
• Scikit-learn

Install manually (if needed):

pip install optuna scikit-learn

---

⚡ Troubleshooting

Problem	Possible Cause	Solution
ModuleNotFoundError	You didn’t install the package	Run pip install -e . in your project folder
optuna not found	Optuna not installed	Run pip install optuna
no attribute run_automl	Wrong import	Use from automl_optuna_demo.main import run_automl
Very low accuracy	Too few trials	Increase trials value in run_automl(trials=50)

---

🧑‍💻 Future Improvements

• Add command-line interface (CLI) for easy execution
• Add support for more models (e.g., SVM, XGBoost, Logistic Regression)
• Visualize Optuna optimization history
• Allow loading custom datasets instead of Iris

---

🪪 License

MIT License © 2025 Vaishnav Naik
Feel free to use and modify this code for learning or projects!

---

💬 Author

Vaishnav Naik
GitHub: [your-github-profile]
Email: [your-email-if-you-want-to-add]

---

🌟 Support

If you like this project, star it on GitHub or share it with others learning Optuna or AutoML!