Why different classifier behaves differently for different data?

r-machine-learning

#1

Why behaviour of different classifier differ for different data?
Based on what parameters we can decide the good classifier for particular dataset?


#2

Hi @nileshthakkar

We have multiple classifiers, each having certain pros and cons. Definitely the performance of classifiers vary for different datasets.

There is no defined way to determine the best classifier beforehand. You’d need to understand the dataset and also the expected outcome. Based on which you can decide which classifiers will be the best for that case.


#3

Hello,

Different classifiers behave differently on the same dataset, but when you compare them on a big set of different dataset some perform better consistently. I encourage you to read this paper that describes it:

http://jmlr.csail.mit.edu/papers/v15/delgado14a.html

And besides that, another very fruitful source of knowledge is what in a very pionnering way Szilard did with his benchmark:

Regards,
Carlos.


#4

Every classifier or machine learning model in general are developed based on certain assumptions.
More importantly ,these assumptions are based on the characteristics of the data set.
so it can be said that, model have inductive bias associated with them.
Classifiers perform best on those datasets which follow these base assumptions