Midterm Peer Review

This project aims to predict the result of election by focusing on 12 swing states. You use county level data such as racial composition, age, etc.
Three things I like about the project:
1)You explain everything in details such as convert the string into int on the numerical data.
2)You use a lot of data and try to deal with messy data problem.
3)You use the correlation plot to select features, and present the picture which is easier for us to read.
Three improvements I think the team can make:
1)Maybe you could explain why you choose 0.25 as the correlation benchmark.
2)Maybe you could also use data in 2004 and 2008 to train your model in case of overfitting problem or test the efficiency of the model.
3)Maybe you could give some suggestion based on your result such as what could the team do in order to win the election.