Can the number of Predictors be different for Train and Test data?

1 view (last 30 days)
Can my train data have 6 predictors while test data have only 4?
The 4 predictors in my test data are the same as in my train data, just that 2 extra predictors of the train might not be present in my test data.
It's basic question, but I'm a bit confused. Any suggestion would be helpful thank you.
  2 Comments
Adam Danz
Adam Danz on 3 Jul 2019
Edited: Adam Danz on 3 Jul 2019
Are you doing cross validation? Are you training a neural network? Generally the training and testing data are subsets of the same dataset.
abhisrisai
abhisrisai on 3 Jul 2019
@Adam Danz
Hi,
I'm using the fitcknn methodology. For now, I'm not using NN. Regarding my dataset, My training data is a different table and my test data is a different table. In other words, 2 different .csv files.

Sign in to comment.

Accepted Answer

Sai Bhargav Avula
Sai Bhargav Avula on 17 Jul 2019
No, You can not have different number of predictors for the train and test data. The model you have trained would require all the predictor that is trained on for an accurate prediction.

More Answers (1)

Greg Heath
Greg Heath on 18 Jul 2019
Of course not. The ultimate purpose of training is to create a model that works well on non-training data.
Thank you for formally accepting my answer
Greg

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!