Why divide the training set and the test set before normalization?

First divide the data into training set and test set, and then normalize the data, and then divide the test set and training set , The difference between the two ：

In theory, the data set should be divided first , And then preprocess the training data , And save the preprocessed parameters , Processing test sets with the same parameters .

Because the division of training set and test set is based on the assumption that only the information of training set is known , And think that the test set data is from the future , Unknown . The information of the test set is used if the test set is divided after preprocessing .

To be one, to be one , If you do not unify, you will not unify . With the same distribution , The test is effective .

Some models have the same effect , Like decision trees . Some must be unified , Such as regression analysis

Reference material

Why divide the training set and test set first and then normalize them ？_CDA Q & a community

(8 A private letter / 35 Bar message ) Does the test set and training set need to be normalized in data preprocessing ？ - You know (zhihu.com)