1
$\begingroup$

I am trying to build a regressor for a dataset which gives info about students' school performance and the probability of getting admitted in the University of their choice.

The first 5 observations look like this :

 GRE_Score TOEFL_Score Uni_Rating LOR CGPA Research Chance_of_admit 0 337 118 4 4.5 9.65 1 0.92 1 324 107 4 4.5 8.87 1 0.76 2 316 104 3 3.5 8.00 1 0.72 3 322 110 3 2.5 8.67 1 0.80 4 314 103 2 3.0 8.21 0 0.65

I have build the following regressors so far : linear regressor, knn regressor and a recurrent neural network. ( I will try a few more later. )

So far, in order to chose among my models, I used the "score" method on the test set for the first two regressors ( it returns the $R^2$ for each one of them ) and the "evaluate" method on the test set for the network ( it returns the MeanSquaredError ).

So, keeping in mind that $R^2$ and MeanSquaredError have different formulas, how can I compare my network with the other two models ??

Any help is much appreciated.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

If you up to a predictive model, you look for a model which performs well on the test set and the metric of interest is the mean squared error which indicates by how much you fail to predict $y$ on average. So don't use $R^2$. Just compare all models based on MSE.

Improve this answer
$\endgroup$
2
  • $\begingroup$ I replaced OLS regressor with Ridge regressor.. Ok it is clear that I should use MSE to choose among my final models.. I have one more question now, regarding the tuning of the models : The tuning of the network depends on the MSE.. However, when tuning the Ridge and the kNN regressors, the "cross_val_score" method uses the R^2 coefficient to choose among the models.. Is it ok ? Or should I also change the tuning so that it depends also on the MSE ? $\endgroup$ Commented Jan 18, 2020 at 14:18
  • 1
    $\begingroup$ I would tune on MSE $\endgroup$ Commented Jan 18, 2020 at 19:15

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.