1
$\begingroup$

I create 4 clusters by using k-mean cluster analysis. After that, I try to classify this target variable with random forest in order to confirm if the clusters are properly grouped. Is it possible?

$\endgroup$
1
$\begingroup$

Generally speaking it is possible. Somemore information would be helpful: How do you plan to test/train? Do you want to use k-fold cross validation, or split the data? Are you going to use exactly the same variables as predictors in the Random Forest, that u used for the clusteranalysis? In this case you propably need to be careful with your interpretation since it is circular by nature.

|improve this answer|||||
$\endgroup$
  • $\begingroup$ I am going to use k-fold cross validation and use different predictors. $\endgroup$ – higenberg May 27 '19 at 7:40
  • $\begingroup$ I think you need to consider the purpose of the analysis. What would be your research question / what is the primary aim? When you are using other predictors you could answer the question: is it possible to predict my new variable (4 categories identfied by CA)? Does such a finding validate the clusters you found? It's depended on your field of research and the existing theories within this field. You may also find this link helpful (Check the answer by Sambhavi Dhanabalan) $\endgroup$ – Björn B May 27 '19 at 13:17

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