Feature union and Function returns with Pipelines

I am struggling to get this pipeline to work. I'm working on a text classification problem where I have one binary feature and the other is text(TFIDF vectorized). I wanted to perform Oversampling to one of the classes and hence I'm defining my own method. Here's my trial so far: `

get_text_data = FunctionTransformer(lambda x: x['FinalText'], validate=False)  get_numeric_data = FunctionTransformer(lambda x: x[['boolean']], validate=False)  pipe_svm = Pipeline([      ('features', FeatureUnion([              ('numeric_features', Pipeline([                  ('selector', get_numeric_data)              ])),               ('text_features', Pipeline([                  ('selector', get_text_data),                  ('xtrain', CustomOversampling(X_train['FinalText']))                                 ]))           ])),      ('clf', svm.LinearSVC(class_weight = 'balanced'))  ])        pipe_svm.fit(X_train,y_train)`                def CustomOversampling(input)       .....             return Combinedmatrix,combinedyframe            TypeError: Last step of Pipeline should implement fit or be the string 'passthrough'. '(<157911x10951 sparse matrix of type '<class 'numpy.float64'>'  

https://stackoverflow.com/questions/65838392/feature-union-and-function-returns-with-pipelines January 22, 2021 at 10:06AM