Swing Scope

Inspiration

All election prediction projects have one thing in common is that they are trying to answer the question of who's going to win? But who is ahead in a safe district is known to campaigns, journalists and analysts but they do need to know what's shifting, however. We wanted to create a model which no one was creating a volatility detector, one that would alert us to which counties are about to change their minds.

What it does

Swing Scope analyzes synthetic electoral data at the county level through clustering of counties according to their economic/demographic features and finding out counties that are politically stable or likely to swing. Machine learning algorithms are used for prediction of swing behavior of the counties as well as probability of victory of the candidates, with model performance assessed using metrics such as F1-score, accuracy, Brier score, and AUC, among others, as well as visualizations.

Built With

• geojson
• git
• google-colab
• jupyter
• kaggle-api
• kagglehub
• lightgbm
• matplotlib
• numpy
• pandas
• parquet
• plotly
• python
• pytorch
• scikit-learn
• seaborn
• streamlit