CODEWITHSUNDEEP
scikit-learnkaggle
AlexanderSch
AlexanderSch

Reputation: 157

Sklearn TruncatedSVD() ValueError: n_components must be < n_features

Hi I'am trying to run script for a Kaggle competition.

you can see the whole script here

But when I run this script i get an ValueError

ValueError: n_components must be < n_features; got 1 >= 1

Can somebody tell me please how to find out how many features there are at this point. I don't think it will be usefull when I set n_components to 0. I also read the documentation but I can't solve that issue. Greetz Alex

Upvotes: 2

Views: 4991

Answers (1)

eickenberg
eickenberg

Reputation: 14377

It is highly likely that the shape of your data matrix is wrong: It seems to have only one column. That needs to be fixed. Use a debugger to figure out what goes into the fit method of the TruncatedSVD, or unravel the pipeline and do the steps by hand.

As for the error message, if it is due to a matrix with one column, this makes sense: You can only have maximally as many components as features. Since you are using TruncatedSVD it additionally assumes that you don't want the full feature space, hence the strict inequality.

Upvotes: 1

Related Questions