Question from a student:

efleonard3

I don't know if this is allowed but could I have operations beyond "+": E.g., using multiplication, division and logarithms to fine tune each predictor?
- If so would it be fair to assume proportionality of the predictors and
Am I also allowed to modify the default rate variable so as to create a closely normal distribution of data points? (E.g., log(default_rate) ~ ...)
Could I also use regularization techniques by limiting the size of coefficients to handle multicollinearity? (E.g. Ridge and Lasso regression)

ivonnem10

These are all great questions! For this competition, we are looking for students to implement their model in such a format my_model <- lm(default_rate ~ ..., data = train) . I believe you are allowed to define any relationship between predictors by using other operations than "+".

We know there are many other advanced ways to generate a model with higher predictive power but we ask all students to follow this format: my_model <- lm(default_rate ~ ..., data = train).

dylarm

efleonard3 Here is where Dash explains that it's fine to use *, so I would imagine most other operators would be fine as well, as @ivonnem10 says.