Abstract

Election forecasting is highly dependent on polling data, yet there is literature surrounding the United States political environment that questions the importance of campaigns. This paper extends that theory to exclude polling data, because it will not be available at the time of research to construct election forecasting models. Relying on economic indicators, approval ratings, party occupancy in congress, and other quantifiable metrics that impact and influence a voter's decision, our forecasting will be built from Linear Regressions, Naive Bayes Classifications, and Decision Trees. We will analyze the performance of our models to identify which models are best at predicting an election with these assorted variables. By incorporating these measures in various predictive modeling techniques, we will simulate the 2020 election, and predict the outcome of the 2020 election in a variety of settings.

Advisor

Pasteur, Drew

Department

Mathematics

Publication Date

2019

Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis

Share

COinS
 

© Copyright 2019 Micheas Yimam