minor changes:

- added docstring for _column_contains_text
- multimodal_wine dataset is moved to more appropriate place
- protected funcs of multi_modal.py are now protected

examples/advanced/multimodal_text_num_example.py

@@ -1,4 +1,5 @@
 import os
+from pathlib import Path
 
 from fedot.api.main import Fedot
 
@@ -8,10 +9,9 @@
 from fedot.core.utils import fedot_project_root
 
 
-def run_multi_modal_example(files_path: str, is_visualise=True) -> float:
+def run_multi_modal_example(file_path: str, is_visualise=True) -> float:
     task = Task(TaskTypesEnum.classification)
-
-    path = os.path.join(str(fedot_project_root()), files_path, 'multimodal_wine.csv')
+    path = Path(fedot_project_root(), file_path)
     data = MultiModalData.from_csv(file_path=path, task=task, target_columns='variety', index_col=None)
     fit_data, predict_data = train_test_data_setup(data, shuffle_flag=True, split_ratio=0.7)
 
@@ -31,4 +31,4 @@ def run_multi_modal_example(files_path: str, is_visualise=True) -> float:
 
 
 if __name__ == '__main__':
-    run_multi_modal_example(files_path='examples/data/multimodal_wine', is_visualise=True)
+    run_multi_modal_example(file_path='examples/data/multimodal_wine.csv', is_visualise=True)

fedot/core/data/multi_modal.py

@@ -102,17 +102,17 @@ def from_csv_time_series(task: Task,
                 'Multivariate predict not supported in this function yet.')
         else:
             train_data, _ = \
-                prepare_multimodal_ts_data(dataframe=df,
-                                           features=var_names,
-                                           forecast_length=0)
+                _prepare_multimodal_ts_data(dataframe=df,
+                                            features=var_names,
+                                            forecast_length=0)
 
             if target_column is not None:
                 target = np.array(df[target_column])
             else:
                 target = np.array(df[df.columns[-1]])
 
             # create labels for data sources
-            data_part_transformation_func = partial(array_to_input_data, idx=idx,
+            data_part_transformation_func = partial(_array_to_input_data, idx=idx,
                                                     target_array=target, task=task,
                                                     data_type=DataTypesEnum.ts)
 
@@ -123,6 +123,7 @@ def from_csv_time_series(task: Task,
 
         return input_data
 
+    @staticmethod
     def from_csv(file_path=None,
                  delimiter=',',
                  task: Task = Task(TaskTypesEnum.classification),
@@ -135,7 +136,7 @@ def from_csv(file_path=None,
         :param columns_to_drop: the names of columns that should be dropped
         :param delimiter: the delimiter to separate the columns
         :param task: the task that should be solved with data
-        :param text_columns: name of columns that contain text data
+        :param text_columns: names of columns that contain text data
         :param target_columns: name of target column (last column if empty and no target if None)
         :param index_col: column name or index to use as the Data.idx;
             if None then arrange new unique index
@@ -149,13 +150,13 @@ def from_csv(file_path=None,
         idx = data_frame.index.to_numpy()
 
         if not text_columns:
-            text_columns = define_text_columns(data_frame)
+            text_columns = _define_text_columns(data_frame)
 
-        data_text = prepare_multimodal_text_data(data_frame, text_columns)
+        data_text = _prepare_multimodal_text_data(data_frame, text_columns)
         data_frame_table = data_frame.drop(columns=text_columns)
         table_features, target = process_target_and_features(data_frame_table, target_columns)
 
-        data_part_transformation_func = partial(array_to_input_data, idx=idx,
+        data_part_transformation_func = partial(_array_to_input_data, idx=idx,
                                                 target_array=target, task=task)
 
         # create labels for data sources
@@ -173,29 +174,34 @@ def from_csv(file_path=None,
         return multi_modal_data
 
 
-def define_text_columns(data_frame: pd.DataFrame) -> List[str]:
+def _define_text_columns(data_frame: pd.DataFrame) -> List[str]:
     """
     :param data_frame: pandas dataframe with data
     :return: list of text columns' names
     """
     text_columns = []
-    for column in data_frame.columns:
-        if column_contains_text(data_frame[column]):
-            text_columns.append(column)
+    for column_name in data_frame.columns:
+        if _column_contains_text(data_frame[column_name]):
+            text_columns.append(column_name)
     return text_columns
 
 
-def column_contains_text(column: pd.Series) -> bool:
+def _column_contains_text(column: pd.Series) -> bool:
     """
+    Column contains text if:
+    1. it's not numerical or latent numerical
+    (e.g. ['1.2', '2.3', '3.4', ...] is numerical too)
+    2. fraction of unique values is more than 0.95
+
     :param column: pandas series with data
     :return: True if column contains text
     """
-    if column.dtype == object and not isfloat(column):
+    if column.dtype == object and not _isfloat(column):
         return len(column.unique()) / len(column) > 0.95
     return False
 
 
-def isfloat(column: pd.Series) -> bool:
+def _isfloat(column: pd.Series) -> bool:
     """
     :param column: pandas series with data
     :return: True if column contains only float or nan values
@@ -207,25 +213,24 @@ def isfloat(column: pd.Series) -> bool:
         return False
 
 
-def prepare_multimodal_text_data(dataframe: pd.DataFrame, text_columns: List[str]) -> dict:
+def _prepare_multimodal_text_data(dataframe: pd.DataFrame, text_columns: List[str]) -> dict:
     """ Prepares MultiModal text data in a form of dictionary
 
     :param dataframe: pandas DataFrame to process
-    :param text_columns: list of text columns names
+    :param text_columns: list of text columns' names
 
     :return multimodal_text_data: dictionary with numpy arrays of text data
     """
     multi_modal_text_data = {}
 
-    for column in text_columns:
-
-        text_feature = np.array(dataframe[column])
-        multi_modal_text_data.update({column: text_feature})
+    for column_name in text_columns:
+        text_feature = np.array(dataframe[column_name])
+        multi_modal_text_data.update({column_name: text_feature})
 
     return multi_modal_text_data
 
 
-def prepare_multimodal_ts_data(dataframe: pd.DataFrame, features: list, forecast_length: int) -> dict:
+def _prepare_multimodal_ts_data(dataframe: pd.DataFrame, features: list, forecast_length: int) -> dict:
     """ Prepare MultiModal data for time series forecasting task in a form of
     dictionary
 
@@ -256,11 +261,11 @@ def prepare_multimodal_ts_data(dataframe: pd.DataFrame, features: list, forecast
     return multi_modal_train, multi_modal_test
 
 
-def array_to_input_data(features_array: np.array,
-                        target_array: np.array,
-                        idx: Optional[np.array] = None,
-                        task: Task = Task(TaskTypesEnum.classification),
-                        data_type: DataTypesEnum = DataTypesEnum.table):
+def _array_to_input_data(features_array: np.array,
+                         target_array: np.array,
+                         idx: Optional[np.array] = None,
+                         task: Task = Task(TaskTypesEnum.classification),
+                         data_type: DataTypesEnum = DataTypesEnum.table):
     """
     Transforms numpy array to InputData object
     """