Combining hyperparameters
It's time to combine all the components of this chapter to improve upon the 78% score obtained through cross-validation.
As you know, there is no one-size-fits-all approach to hyperparameter fine-tuning. One approach is to input all hyperparameter ranges with RandomizedSearchCV. A more systematic approach is to tackle hyperparameters one at a time, using the best results for subsequent iterations. All approaches have advantages and limitations. Regardless of strategy, it's essential to try multiple variations and make adjustments when the data comes in.
One hyperparameter at a time
Using a systematic approach, we add one hyperparameter at a time, aggregating results along the way.
n_estimators
Even though the n_estimators value of 2 gave the best result, it's worth trying a range on the grid_search function, which uses cross-validation:
grid_search(params={'n_estimators':[2, 25, 50, 75, 100]})
The output is as...
