0

In my dataset, I explored there are actually three layers of labels. All are not independent labels. For the first layer, the label is binary. If the class is 1, we can go to the further layer. The second layer has four different class and each class can belong to 42 different classes. In a nutshell, the class or labels are in a tree. Now I want to perform a classification to find out all three layers of labels in a single experiment.

This image https://drive.google.com/file/d/1M8yVI7mMBa9Wr9q7UQQZQ-X3q9MBDGaT/view?usp=sharing shows the layers of labels.

I want to classify if a given data is malware or not. If it is malware, I want to find out its category and which family it belongs to. I want to do all these in a single process. I heard about scikit multi-label classification, but I am not sure if it is really a multi-label classification problem.

Nasimul Hasan Chowdhury
  • 15
  • 4
  • You can try multiclass classification, i.e. you output will be flattened. 1D array (Malware, Beign, Adware, Ransomware ....... ) and so on. And for adware both Malware and Adware classes will be 1, and eventually its subclasses will be 1. I haven't tried it before, but I presume it can work. – venkata krishnan Feb 03 '20 at 07:47
  • @venkatakrishnan Can you please refer me to an example from where I can get the idea clearly you described? – Nasimul Hasan Chowdhury Feb 03 '20 at 10:45
  • https://towardsdatascience.com/multi-label-classification-and-class-activation-map-on-fashion-mnist-1454f09f5925 this is using fashion mnist for multiclass classification – venkata krishnan Feb 04 '20 at 01:49

0 Answers0