Which is the (single) model that you get after k-fold cross-validation

Question

Let's assume we perform a ML based classification in terms of some supervised ML approach. In view of a bias-variance trade-off, we decide to use the k-fold cross-validation strategy. For each fold we get a confusion matrix and a classification report ...eventually a ROC-curve. Thus we get k such results. One might characterize the overall score of each such fold by the mean-squared error (MSE). Thus, we get k times different MSE scores. One might take the average of the MSE's and find the mean score of the model based on the k-fold cross-validation, MSE_mean.

Here is the code that I run in the case of SVM:

from sklearn.model_selection import KFold
from sklearn.utils import shuffle
from sklearn.svm import SVC
from sklearn.metrics import classification_report, confusion_matrix

def SVM(X_train, X_test, y_train, y_test, kernelType):
    svclassifier = SVC(kernel=kernelType)
    svclassifier.fit(X_train, y_train)
    y_pred = svclassifier.predict(X_test)
    confusionMatrix = confusion_matrix(y_test,y_pred)
    classificationReport = classification_report(y_test,y_pred)
    return confusionMatrix, classificationReport
xall, yall = shuffle(x, y, random_state=21)

kf = KFold(n_splits=5)
from imblearn.over_sampling import SMOTE
for train_index, test_index in kf.split(x):
   yall.iloc[train], yall.iloc[test]
   x_train = xall.iloc[train_index,:]
   y_train = yall.iloc[train_index]
   x_test = xall.iloc[test_index,:]
   y_test = yall.iloc[test_index]

 # smote
smote = SMOTE(sampling_strategy='minority')
x_train_sm, y_train_sm = smote.fit_sample(x_train, y_train)
x_train_sm, y_train_sm = shuffle(x_train_sm, y_train_sm, random_state=21)


#confusionMatrixSVM, classificationReportSVM 
confusionMatrixSVM, classificationReportSVM = SVM(x_train_sm, x_test, y_train_sm,   y_test, 'rbf')
print(confusionMatrixSVM)
print(classificationReportSVM)

My questions: Since we do ML in order to make predictions, we want to have a single model at the end of the k-fold cross-validation process and not separate k models. How do we get a single model at the end ? That model should in principle provide a score MSE = MSE_mean. I need that model for making predictions. Can you suggest some code for getting a single model after the k-fold cross-validation ?

Many thanks.