In more detail, if the values of the feature-set is already between 0 and 0.65, do I really need to normalise/standardise the data before I apply machine learning algorithm of regression?
I would suggest you to study data distribution at start, as normalization/standarization will deformate data distribution and thus would alter results performance. Depending on data origin, you can apply the preprocessing employed in the literature, as performance will definitively depend on data preprocessing.
I think you have to normalize data after regression, because objective functions will not work properly without normalization (re-scale to [0 -1] or [-1-1], normalize to zero mean).
Whether you want to normalize or not will depend on the relative ranges of the features in your feature set and how sensitive the algorithm you are using is to these differences. For example, if you have two features A and B where A is in the range of 0-1 and B is in the range of 0-0.01 then you may want to normalize them to both be in the same range.
I guess it depend on how your data is.
Since it is a research, no one know how will it be the best solution. And at final, you have to try all possible approaches.
Good luck
- Badges
- Science topic
- Similar topics
- Social Psychology
- Values