0
$\begingroup$

I've recently embarked on my data science journey, and I've therefore also started a data science course.

In this course, we've received an assignment asking us to model a data set using different supervised algorithm (logistic regression, SVM, classification trees, random forest).

Once we've built the models, we're asked to compare them. I get the theoretical pros and cons (blackbox, accuracy, etc.).

My question, however, is related to comparing logistic regression to the remaining algorithms. In the remaining algorithm I get an accuracy through classification tables as I've utilized training/testing sets. These are easily comparable. This has not been done with the logistic regression, as I've not been taught how to in the course.

So, my question is: How do i compare logistic regression to the remaining supervised algorithms?

Thanks in advance.

Improve this question
$\endgroup$
1
  • $\begingroup$ I assume you are talking about 2-class problems. I have never heard the term "classification tables" but I assume it is a table that tells you which objects belong to class 0 and which ones belong to class 1. Now it seems your problem is that you get no class label from logistic regression but a value between 0 and 1 for each object. Usually you assign all objects with a value of < 0.5 to class 0 and all objects with value >0.5 to class 1 and proceed as with other classifiers. Please clarify in case this does not describe your problem. $\endgroup$
    – SaiBot
    Oct 31, 2018 at 8:16

1 Answer 1

Reset to default
0
$\begingroup$

I assume this is a classification problem. You should be able to create a classification table with the output of logistic regression as well, and once you have that you can compute the accuracy for all of the algorithms. To create a classification table you have to count the True Positive, True Negative, False Positive, and False Negative cases. You can look at this existing thread for tips on how to do it in R: Classification table for logistic regression in R

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.