Supervised Learning: Donor Classification

Summary:

In this project, I help CharityML maximize the likelihood of receiving donations through constructing a model that predicts if a person receives income exceeding 50k/yr; a level known to indicate being a good candidate for donations.

• Full writeup with code
• Full writeup without code
• Slides without code

Project Organization

1. Exploratory Data Analysis
2. Data Engineering
3. Metrics
4. Machine Learning Models
5. Summary

Models

• Naive Bayes
• Logistic Regression
• Random Forest
• AdaBoost
• Gradient Boost
• Extreme Gradient Boost
• K-Nearest Neighbors

Performance

• Accuracy: 87.26%
• F-0.5 Score: 76.05% (high precision model)

Technologies

Python inside of an IPython Notebook and published with Reveal.js
Employed libraries:

• Pandas==1.0.1
• Numpy==1.18.1
• Scikit-learn==0.22.1
• Scipy==1.4.1
• Statsmodels==0.11.0
• Plotly==4.6.0
• Seaborn==0.10.0
• Matplotlib==3.1.3
• dython==0.5.0.post2
• Custom Module: visualization.py
• Custom Module: modeling.py
• Custom Module: visuals.py