Conducted predictive classification modelling and performance evaluation for several models used to predict the political affiliation (target variable) of random U.S citizens.
-
Administered EDA using Tidyverse and ggplot2 in R Studio to discover the best predictors of political affiliation with regards to correlation on the training dataset
-
Built, trained, tuned and tested classification models including logistic, KNN, random forests, LDA, QDA and SVM, to best predict the political affiliation of a random U.S citizen in terms of accuracy
-
Achieved an accuracy of 65.08 % using a tuned LDA model on the validation dataset (top 20% in the world)