This project applies machine learning techniques to predict nightly prices of Airbnb listings and classify listing categories starting from a synthetic dataset. It includes a a Jupyter notebook for EDA and Python scripts that cover the complete data science workflow from exploratory data analysis (EDA) to model training, tuning, and evaluation.

Project Structure

• EDA.ipynb: Exploratory Data Analysis to uncover insights from the data.
• regression.py: Scripts to train, tune, and evaluate regression models
• classification.py: Scripts to train, tune, and evaluate classification models
• nn.py: script for training the neural networks
• utils.py: Utility functions for data handling and model management.
• models/: Directory containing trained models and their metrics for easy access and reproducibility.

• Python
• NumPy
• Pandas
• PyTorch
• Scikit-learn
• XGBoost
• TensorBoard

• Logistic Regression
• Decision Trees
• Random Forest
• Gradient Boosting
• XGBoost

• Linear Regression
• Decision Trees
• Random Forest
• Gradient Boosting

• Custom PyTorch implementation

• Comprehensive model comparison for both classification and regression tasks
• Hyperparameter tuning
• Model evaluation and metrics tracking
• Neural network implementation with PyTorch
• TensorBoard integration for visualizing training progress

1. Clone this repository
2. Install dependencies: pip install -r requirements.txt
3. Run the Jupyter notebooks or Python scripts to train and evaluate models

While the performance metrics may not be impressive due to the artificial nature of the dataset, this project demonstrates the ability to:

• Implement various machine learning algorithms
• Perform data preprocessing and feature engineering
• Conduct model evaluation and comparison
• Utilize popular data science and machine learning libraries

For detailed results and visualizations, please refer to the Jupyter notebooks

• Apply these techniques to real-world datasets
• Implement more advanced feature engineering techniques
• Explore ensemble methods and model stacking
• Extend the neural network architecture for more complex tasks