How to optimize multiclass classification model for recall

Question

I am working on a churn model with multiclass GBM classification. I am using the below code but not sure what I can do to maximize recall for classes 0 & 1. Would changing the order of classes help?

                precision   recall  f1-score   support

           0       0.67      0.11      0.19      1035
           1       0.90      0.03      0.06      2663
           2       0.84      1.00      0.92     18835

    accuracy                           0.84     22533
   macro avg       0.80      0.38      0.39     22533
weighted avg       0.84      0.84      0.78     22533

Code

param_test1 = {'max_depth':(3,5)}
estimator = GridSearchCV(estimator=GradientBoostingClassifier(learning_rate = 0.05, loss='deviance',subsample=0.8,random_state=10,
                                                           n_estimators=200),param_grid=param_test1,cv=2)

estimator.fit(df[predictors],df[target])