Normalization in Linear Regression

Normalization in Linear Regression
0

Hi,

I’ve read normalization should be used whenever distance matrix is used in the algorithm (like linear regression). But what is the necessity as anyway coefficients would be adjusted for different range of inputs. Accuracy would remain same in both the cases (with and without normalization)? The only effect would be the speed of training (faster for with normalization?)?

Hi,
Yes you’re right at some point, but in data science when we don’t rely on training models but the simpler statistical analysis, Normalization becomes an essential aspect.
When we do simple exploration of data, there are times when we need to normalize data.

1 Like

Hi ,
Normalization or standardization are performed to bring all features data to a standard scale. For example if many university students are applying to B-school admission and all of these universities evaluates on a different scale then it would be hard to measure their abilities if we do not standardize their scores.

Standardization and normalization should be performed wherever we are dealing with mathematical equations. In linear regression we use gradient dscent method for learning and loss function is also a continuous value we must standardize the feature values in order to obtain good accuracy but for decision trees we don’t have to do this because there are no math involved in decision tree apart from many if and else.

2 Likes