3

I am using Weka to train a model from few days. I know Weka use Java code to implement a classifier. I also heard that Weka has some github pages to describe the java code for the classifiers. I like to know the SVM java code which is used in WEKA. I found few webpages describing Java code for SVM classifiers for WEKA. I can not understand which one is there official page.

Providing me the link to SVM GitHub page would be very helpful.

Thank you.

Encipher
  • 359
  • 1
  • 9

1 Answers1

5

The official Weka source code is stored in a local Gitlab git repository (not on Github).

Note that there are two versions of SVM commonly used with Weka:

Erwan
  • 24,823
  • 3
  • 13
  • 34
  • I am looking for the code to understand how categorical data is handled in SVM. Because I found if I feed the SVM classifier with categorical data classifier did not throw the error, which is common for sklearn. So I want to know what part of the java code categorical data has been handled. – Encipher Oct 08 '22 at 22:20
  • @Encipher in Weka categorical features must be declared as `nominal` in the `arff` file, this is how it's handled (not specific to SVM). I assume that any feature declared as `nominal` is one-hot-encoded, but I never actually checked the code. – Erwan Oct 09 '22 at 20:21
  • Yeah I searched and found they have an unsupervised filter `nominaltobanry` which is based on one hot encoding and they just import that file on SVM. – Encipher Oct 10 '22 at 20:08