Now lets see the predictions
Let's try again considering now the number of comments.
Here $$\theta_1$$ changes slightly due to the fact that there is a new variable ($$\theta_2$$) involved in the equation. This aftects the minimization of the MSE by whatever method the regression uses. If the regresor uses gradient descent, then the direction of descent will change and so will the minimum values.
Now lets train a polinomial fit.
The problem relies on the fitting since we're using the same set for training and testing.
A split of train and validation should improve it a bit.
Let's load the data
Now lets fit a logistic regressor to the data
We can see that the regressor is quite good with the test data, but that is not surprising since it was trained with it.
This makes no surprise as the data is primarily Male so theres a good chance of picking a Male. Let's train it with balanced weight now.
An easy way to improve the model is to consider the ABV and a one hot encoding of the beer styles.
Now lets train it