https://github.com/xprithvi/random-forest-regressor

This Jupyter notebook serves as a machine learning template to quickly make predictions and analyse feature importance in a dataset.
https://github.com/xprithvi/random-forest-regressor

data-science feature-extraction machine-learning random-forest random-forest-regression scikit-learn

Last synced: 21 days ago
JSON representation

This Jupyter notebook serves as a machine learning template to quickly make predictions and analyse feature importance in a dataset.

• Host: GitHub
• URL: https://github.com/xprithvi/random-forest-regressor
• Owner: xPrithvi
• License: mit
• Created: 2023-09-20T11:32:20.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-09-20T19:06:37.000Z (over 1 year ago)
• Last Synced: 2024-11-19T16:59:37.162Z (3 months ago)
• Topics: data-science, feature-extraction, machine-learning, random-forest, random-forest-regression, scikit-learn
• Language: Jupyter Notebook
• Homepage:
• Size: 2.35 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Random Forest Regressor

This Jupyter notebook serves as part of the data science pipeline by providing a quick and easy framework to

perform feature enginnering, model training and feature importance analysis for data exploration. In this particular notebook,

Sci-Kit Learn's RandomForestRegressor was trained on information regarding [housing in Perth](https://www.kaggle.com/datasets/syuzai/perth-house-prices) to

numerically predict house prices based on floor space, suburb, number of bedrooms, etc. Feature importance analysis was performed using 

built-in methods that calculate importance by node impurity. However, SHAP was also used to provide a more robust and in-depth analysis

via Shapley values.

## Features

- Model saving and loading.

- Hyperparameter tuning via Bayesian optimization.

- Feature importance analysis using tree node impurity and Shapley values.

## Future Improvements

- Custom user input to the model (involves writting a custom data encoder instead of using pandas.get_dummies()).

- Reducing the disk size of saved models.