I am working on clustering algorithms. I am working with titanic dataset. It contains 6 categorical features. I used k-means algorithm on this dataset. I am using label encoding for categorical features. But I found that categorical features should use euclidean distance. It should use Hamming distance. So, how to make k-means work finely on mixed features? I don't need other algorithm. I just want to work with k-means only on mixed features dataset.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ You should use one-hot enconding rather than label encoding $\endgroup$– ChopinCommented Jun 23, 2020 at 12:16
-
$\begingroup$ How one-hot encoding solves the problem? since using one-hot encoding also results 0s and 1s.when i apply k-means it calculates distance.so it may results in decimal .How to solve that. @Chopin $\endgroup$– SathyaCommented Jun 24, 2020 at 15:51
Add a comment
|
4 Answers
$\begingroup$
$\endgroup$
Label encoding is not a good idea if the nature of categories are not ordinal (it is actually not my favorite anyways). Use one-hot encoding and see how it works. You may apply a feature extraction on top of it, e.g. PCA, to reduce the noise coming from sparsity. The other idea is to label categories by their fraction in the feature, for example:
[a,b,b,c,a,a] --> [3/6, 2/6, 2/6, 1/6, 3/6, 3/6]
answered Jun 24, 2020 at 10:24
$\begingroup$
$\endgroup$
The best way to encode the data will be through any encoding mechanism like label encoder etc. But before handling the categorical variable check the correlation of a categorical variable with the target variable using the feature selection methods like chi square test with selectKbest.
$\begingroup$
$\endgroup$
You can quantify correlation, or more precisely association, between categorical variables using something like cross-entropy. There’s an available library dython to compute such association values. Also I am curious why do you want to do clustering ? What is your expected output?
$\begingroup$
$\endgroup$
0
I think the k-prototype algorithm is what you are looking for.
