Question

(can have zero or multiple answers)

Transformation of training data representation...

1. is sometimes done before cluster modeling

2. is sometimes done before predictive modeling

3. can include variable selection

4. can include adding synthetic variables

5. can include imputing missing values based on variable means

6. can include imputing missing values to be zero

7. can affect predictive model performance

8. does affect how testing data must be represented

Answer 1

Answer:-

Both 1 and 2 are correct regarding Transformation of training data. For both clustering and prediction we first transform training data appropriately using normalization , discretization and many techniques.

3 and 4 also correct. data transformation includes selecting appropriate varibles and adding synthetic data generated from existing data like in PCA.

5 and 6 do not comes under data transformation it comes under data clearning.

7 and 8 are correct. Data transformation effects how our model performs by effecting variance in data and other things. For predictive model it effects performance by chanding range of data like in between 0 an 1.

Yes it effects testing also . While performing testing we must follow same procedute followed by transformation on training data to make testing data compatible with trained model.

Thank you....