Pondering over “Multi Label” vs “Multi Class” Classification? This is for you…

While working mostly with “Multi Class” classification, got to hear word “Multi-Label” classification.

 First thought struck my mind, “might be just another name for Multi Class”.

My attempt to dig it down and work over “Multi-Label” classification problem pushed me to take this to larger audience. So that people no more juggle between these two words.

Let’s start….

Figure 1:

Keeping above image in mind, lets jot down the relevant things that come to our mind:

So, considering above listed things as labels, a village can be represented by multiple labels.

Figure 2:

Keeping above image in mind, lets jot down Yes/No for the same set of labels, which depicts a village.

So, one village can be represented by more than one label. Hence it is a Multi-Label Classification. Where we have set of target variable. And an entity can fall under more than one target variable.

However, one village at a time can belong to either North-India, South-India, West-India or East-India. This is a type of multi-class classification problem. One village can only be classified under one category.

Facilitator:

”Scikit-multilearn” is a library which facilitates multi-label classification and that is built on top of the well-known “scikit-learn” library.