how to find important variable using knn in R

问题

I want to upgrade accuracy using KNN algorithm.

I have 23 factors(sex, age, payment, education etc...)

Problem is there are too many variables so i want to know which are effective variables.

[info]

• dataset -> 10000 row, 24 column

• last column is default(1=yes, 0=no)

I divided 7000 for training set and 3000 for test set. When i use all variables, i got error classification about 1000. Also, using ROC Curve shows 800 error. But i want to reduce error rate more. What method i can do?? If you write code example, it will be helpful!! :)

来源：https://stackoverflow.com/questions/52878066/how-to-find-important-variable-using-knn-in-r