Complete this project to continue the work we did with the Kaggle house price dataset.
Using the network from the video, choose 2 more categorical columns from the house price dataset, and incorporate those into the neural network. Does this result in better accuracy? (Remember, that lower % is better in our custom accuracy function).
Now try to make the choice of categorical data columns generic - so that you can just define an array of categorical columns to use, and your function will create and use the embeddings for you.
Using your more generic form of categorical columns, choose every numerical and categorical column that you can find in the house price dataset, and use that to train a new network. What is the best accuracy that you can get?