1
$\begingroup$

I'm attempting to split my data set into 70% training, 15% testing and 15% validation.

train_X, test_X, train_Y, test_Y = train_test_split(data, labels, test_size=0.3, train_size=0.7,random_state=1,stratify = labels)

test_X, val_X, test_Y, val_Y = train_test_split(test_X, test_Y, test_size=0.5,
                                                    random_state=1,stratify = labels)

But I'm not sure if this code splits the test set into half. Also, I keep getting this error:

     29 def main():
     30     data, labels = load_data()
---> 31     train_X, train_Y, val_X, val_Y, test_X, test_Y = process_data(data, labels)
     32 
     33     best_model, best_k = select_knn_model(train_X, val_X, train_Y, val_Y)

/tmp/ipykernel_50/3409802801.py in process_data(data, labels)
     45     X_counts = vectorizer.fit_transform(train_X)
     46     X_count = vectorizer.transform(test_X)
---> 47     Xval = vectorizer.transform(Val_X)
     48     # Return the training, validation, and test set inputs and labels
     49 

NameError: name 'Val_X' is not defined

How do I fix this?

$\endgroup$
1

1 Answer 1

1
$\begingroup$

Do not split the test set into half for the second train_test_split. Instead first split your whole data into train and test set. Then split the train set into train and validation sets as shown below.

X_train, X_test, y_train, y_test 
    = train_test_split(X, y, test_size=0.2, random_state=1)

 X_train, X_val, y_train, y_val 
    = train_test_split(X_train, y_train, test_size=0.25, random_state=1) # 0.25 x 0.8 = 0.2

Regarding the error, you have defined val_X in your second split but you are using Val_X when using the vectorizer. Just correct the uppercase into lowercase and you should be fine!

$\endgroup$
1
  • $\begingroup$ I think all ML flow are like this, prepare original data-->prepare train data, prepare test data.. After that experiments, building models, fitting models should be conducted on train data not on test data, hence it is good practice to create validation data from train data. Next, try to evaluate your model on test data. However note that validation data is required if you want to avoid overfitting, improve performance, or break model fitting between epochs, . For more details refer , datascience.stackexchange.com/questions/15135/… $\endgroup$
    – niran
    Commented Jan 27, 2023 at 15:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.