Polynomial Regression & Machine Learning

Finally, we have worked through the entire machine learning algorithm, which can be summed up like this:

i$.$ Randomly split the data into train and test sets

ii$.$ Fit polynomial models of varying degrees on the train set

iii. Select the model that minimizes average error on the test set. Use this model to make predictions on new data.

iv$.$ As you collect more data, automatically repeat the above steps $($to refine your model$)$

Using this algorithm, in what ways did the machine "learn" the relationship between income and produce buying? In what ways is the model still imperfect? Explain.