Suppose I have a table with some factor characteristics of some plants. For instance, petal color, pollen color, and so on. What is the best way to classify that data? Is it feasible to use some of these methods?

  • Discriminant Analysis LDA, QDA.
  • KNN (k Nearest Neighbours).
  • Naive Bayes

EDIT: Suppose I have two groups of plants - the ones that are edible, and the others that are not. Now, I want to apply some classification method, that given the properties of a new plant, to classify it as edible or not.

  • 1
    $\begingroup$ To clarify the question: are you trying to find natural groupings (previously not known or considered) based on several measured variables for the plants? $\endgroup$ – LSC Apr 14 '19 at 12:00
  • $\begingroup$ Why should this be a problem for any classification algorithm? $\endgroup$ – Tim Apr 14 '19 at 13:45
  • $\begingroup$ @LSC I have edited my post $\endgroup$ – alienflow Apr 14 '19 at 14:19

Just use dummy variables for your categorical variables. Then, for the edible/nonedible example, you could try logistic regression, which will output a probability, which is more informative than simply a classification. A probability close to half would indicate a sample that is difficult to classify.

For more than two levels, you could try multinomial logistic regression, but many other linear methods would also accept dummy variables.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.