Different models have different data requirements. Some steps may or may not be needed.
Generally, simple models such as tree model and LR are used.
Tree model
Due to the power of lgb, there is no need for normalization, data bucketing, category onehot, etc.
After the feature is constructed, the data file can be saved
Simple model
Such as Ridge and LR, you need to perform data bucketing, normalization, and category onehot