How to pass an argument to a function within a customized function?

Question

First of all the code snippet:

## Packages
from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.metrics import fbeta_score
from imblearn.over_sampling import RandomOverSampler
from sklearn.datasets import make_classification 

## Create dataset

class_weight = list([0.90])

X, Y = make_classification(n_samples = 1000, n_classes = 2, n_clusters_per_class = 2, weights = 0.9,
                         n_features = 10, n_informative = 10, class_sep = 1, shuffle = True)

X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size = 0.2, stratify = Y)

## Define Function

def resampled_fit(X, y, model, model_params, sampling, sampling_params):

    sampling = sampling(**sampling_params)    ## <= here is the issue

    X_train_balanced, Y_train_balanced = sampling.fit_resample(X, y)

    # Fit the model to the balanced training data
    model_obj = model(**model_params).fit(X_train_balanced, Y_train_balanced)

    # Compute performance metrics

    Y_pred = model_obj.predict(X_train_balanced)

    f2_val = fbeta_score(Y_train_balanced, Y_pred, beta = 2)

    return f2_val

## Define Inputs

ROS = RandomOverSampler(random_state = 42)
ROS_params = dict(sampling_strategy = 0.8)

SVM = SVC()
SVM_params = dict(kernel = 'rbf', probability = True)

output = resampled_fit(X, Y, SVM, SVM_params, ROS, ROS_params)

Basically I want to input 'sampling_strategy' as a seperate argument (for RandomOverSampler) in my customized function in the same manner as I input the parameters for the classifier object seperately. However it does not work like this.

I get the error message:

TypeError: 'RandomOverSampler' object is not callable

I checked the type of the RandomOverSampler function but it is abc.ABCMeta in the same way as the classifier object. What is the workaround to overwrite the input arguments of the RandomOverSampler within the function?

PS: Yes, I need to input the arguments seperately because I want to optimize the function with grid search afterwards. Obviously you need to do CV to use balanced sampling but as already mentioned this is only a snippet.

I would be thankful for any help regarding this issue.

score 0 · Answer 1 · answered May 25 '20 at 08:58

Problem fixed by inputting the functions 'RandomOverSampler' and 'SVC' without brackets.

If the input arguments are delivered with the function, then we can use the brackets, e.g.

RandomOverSampler(random_state = 42)

if we wish to input the arguments for it seperately, then we have to input the function as

RandomOverSampler

.

How to pass an argument to a function within a customized function?

1 Answers1