sklearn cross_val_score and cross_validation notes

[Ritchie Poon]

Compare major properties:

i  Fit model: Yes  / Yes

ii. Return fitted model:   No   / Default No, yes if change default, 

but only model for individual fold

iii Return scores: Yes / Yes test + train score

In general,  you first use Cross Validation to assess (score) different model performance (with different hyperparameters), then re-train (fit) a model (based on the set 

of best performing hyperparameters).

You can also use Grid Search (GridSearchCV) which will  walk through different parameter grids and evaluate using CV to select the optimal set of parameters.

So cross validation is an evaluation tool for a given  model +hyperparameters to avoid overfitting on a limited set of data

machine learning - sckit-learn Cross validation and model retrain - Data Science Stack Exchange

machine learning - What is the difference between cross_validate and cross_val_score? - Data Science Stack Exchange

python - Reusing model fitted by cross_val_score in sklearn using joblib - Stack Overflow