My data set has two features Date and Name, which is not adding any value to Target Variable. Should I completely drop these two features before constructing Train and Test Data sets, will this drop improves accuracy?
If any feature doesn't add any value to the Target variable it is best to drop it from the dataset. But you shouldn't jump to conclusions about any feature without once comparing the performance of the model with and without the feature. If performance improves after dropping the feature, then it's fine, but if it doesn't then that feature might have some value and shouldn't be dropped without further investigation.
You can drop a feature if it consists entirely of unique values, such as the ID column in a dataset that will have unique values for each row or some other feature like this which has very high cardinality (meaning unique values).
Comments
0 comments
Article is closed for comments.