Predictive Analysis of United States Presidential Elections Using K-Prototype Clustering

Abstract

This project aims to create a predictive modeling tool that can be used in order to forecast future major United States political elections. In other words, I will create a K-Prototype clustering model that will predict which way a state will vote during an election. I will optimize the results of this clustering model utilizing various cost estimation methods which will grant me unique insights into the United States political sphere. In order to do so, I will obtain relevant and reliable data from various reputable sources and use data manipulation techniques in order to clean the gathered information. From this, I plan on creating a predictive modeling tool that will forecast future elections based on randomized data. At a minimum, this project will identify key clusters of voter classifications as well as determine relevant identifiers that strongly influence the outcome of election results.