Custom CV in RecursiveFeatureAddition, list index out of range - cross-validation

Question

I have the following problem
'List index out of range'
Here my code:
rfa2_search = RecursiveFeatureAddition(mod_XG, scoring=accuracy, cv=folds, threshold=0.01)
folds variable is obtained through this split:
def walk_forward(df_train, n_fold, zip_):
# non-overlapping walk forward
total_points = df_train.shape[0] # 5000
sub_val_points = int(0.5 * total_points / n_fold) # 500
sub_train_points = int(total_points - sub_val_points * n_fold) # 2500
arr_tra, arr_val = [], []
step = sub_val_points # non overlapping criteria
pos = 0
for _ in range(n_fold):
arr_tra.append(np.arange(pos, pos + sub_train_points))
arr_val.append(np.arange(pos + sub_train_points, pos + sub_train_points + sub_val_points))
pos += step
if zip_:
custom_cv = zip(list(arr_tra), list(arr_val))
else:
custom_cv = [list(arr_tra), list(arr_val)]
return custom_cv
I'm sure, SURE, that the index that are stored in the zip object is correct with X_train, what I'm saying is that there is NO index in folds that are out of shape of X_train (last index is 4535 and X_train.shape[0] is 4536, so it is correct). However, if I run
rfa2_search.fit(X_train, y_train)
it gets the error:
IndexError: list index out of range
If I run with an integer value in cv (for ex cv=5), the call of .fit method will not produce the error so the error is in folds = walk_forward(..) but I'm sure (I print and check manually) that the indexes obtained with walk_forward are compatible.
Moreover, I've already used the folds from walk_forward as a parameter for GridSearchCV and it works!!!
So the problem, imho, is of RFA implemented
What can I do for solving this issue?

Answer 1

I changed the function as follows:
def walk_forward(df_train, n_fold, zip_, skl):
total_points = df_train.shape[0] # 5000
sub_val_points = int(0.5 * total_points / n_fold) # 500
sub_train_points = int(total_points - sub_val_points * n_fold) # 2500
custom_cv, arr_tra, arr_val = [], [], []
step = sub_val_points # non overlapping criteria
pos = 0
if skl:
for _ in range(n_fold):
arr_tra.append(np.arange(pos, pos + sub_train_points))
arr_val.append(np.arange(pos + sub_train_points, pos + sub_train_points + sub_val_points))
pos += step
if zip_:
return zip(arr_tra, arr_val)
else:
return [arr_tra, arr_val]
else:
for _ in range(n_fold):
arr_tra = np.arange(pos, pos + sub_train_points)
arr_val = np.arange(pos + sub_train_points, pos + sub_train_points + sub_val_points)
pos += step
arr = (arr_tra, arr_val)
custom_cv.append(arr)
if zip_:
return zip(custom_cv)
else:
return custom_cv
The was in the documentation of feature_engine libraries: it accepts as a custom fold the same way of one obtained with Kfold.
I've discovered this after infinitely try, hope someone find this helpful!