Final Peer Review #17
Description
The project deals with predicting the outcome of the 2020 US elections based on the voting rates in key swing states - that often determine the result of the election. The data they collected included information on racial composition, education levels, health care levels, age, gender, etc. What was interesting that they were able to utilize the actual results from the elections to compare their model performance.
The project has done a great job in handling the data. A very thorough data cleaning procedure, followed by feature selections and transformations. These were very well supported by relevant visualizations. I also appreciate the breakdown of features. I appreciate the honesty about the algorithmic bias that can creep into your models. And lastly, great work on bringing in data and providing insights from the 2020 elections, especially the whole section for mail in ballots. It really helped tie in your project well and made for an interesting read.
Areas to develop the project could be using different models. Lin Reg works well for the kind of data but it would be interesting to explore any hidden non linearities in the data with more complex models. Furthermore, I think it would be interesting to have seen the predicted results from individual counties (BLUE or RED). Phrasing it as a classification would have provided other metrics (F1, Recall) to explore and further evaluate performance and reliability of the models. Overall a good project and very relevant project!