economic-presidential

This repo contains 2 projects exploring the relationship between economic indicators and results of U.S. Presidential elections from 1976 to 2020, while using using Python, Jupyter Notebook, scikit-learn, pandas, geopandas, numpy, scipy, matplotlib, seaborn.

Note: feature-selection-linear-regression was completed in September 2023, 2020-US-presid-election-by-state-classification was completed in January 2024.

Brief summary of the projects

• Data cleaning, data wrangling, creating a custom dataset
• Feature engineering
• Linear Regresssion models
• Feature selection with Lasso Regression
• Logistic Regression models
• Support Vector Machines (with and without kernels)
• Random Forest models
• Gradient Boosting models
• Ensemble learning
• Model explanations: Gini importance, permutation feature importance, LIME.
• Choropleths (geographic plots).
• High scores achieved predicting 2020 election results on 1976-2016 data (see below).

Data summary

• U.S. President 1976–2020 - state-level results of U.S. Presidential electionsin 1976-2020 by MIT Election Data and Science Lab.
• State-level economic data - historical data on employment, income, taxes, benefits etc., by U.S. Bureau of Economic Analysis.

feature-selection-linear-regression

• Data cleaning code
• Feature engineering code
• Resulting dataset
• Linear regression models and detained project summary

2020-US-presid-election-by-state-classification

• Feature enginnering for classification
• Code for all classification models, choropleths and model explanations
• Full report on the project