Feature:: Preprocessing models (#104)

* add `Binarizer` in the `SKLEARN_PREPROCESSING_TABLE`

* add preprocessing models' tests runner

* add `Binarizer` test

* `CHANGELOG.md` updated

* `SUPPORTED_MODELS.md` updated

* fix : index error in `SUPPORTED_MODELS.md` fixed.

* update : extra lines adde to `binarizer.py` added.

* add : test for `OneHotEncoder` added.

* fix : `toarray()` added to `one_hot_encoder.py`.

* fix : trailing white spaces removed.

* add : tests added for `LabelBinarizer`.

* add : tests for `LabelEncoder` added.

* add : tests added for `StandardScaler`.

* update : preprocessing tests updated.

* add : `FunctionTransformer` preprocessing model added.

* .gitignore updated

* functionTransporter enhanced to support numpy ufunc

* preprocessings test dedicated utils added

* storage directory added containing preprocessings exported files in json

* add full end to end transporting to preprocessing modules tests

* add FunctionTransporter to Preprocessing chain

* fix : codacy issued fixed.

* add : `KBinsDiscretizer` preprocessing model added.

* fix[?] : nd arrays stack in the general data structure.

* [revert] : `KBinsDiscretizer` reverted temporary.

* add : `KernelCenterer` preprocessing model added.

* update : minor updates in preprocessing tests.

* add : `MultiLabelBinarizer` preprocessing model

* add : `MaxAbsScaler` preprocessing module added.

* add : `Normalizer` preprocessing model added.

* add : `OrdinalEncoder` preprocessing added. [:skull:] check the Nan value in json for parsing.

* fix : codacy issues fixed.

* add : `PolynomialFeatures` preprocessing model added.

* add : `PowerTransformer` preprocessing model added.

* fix : typo fixed for `RobustScaler`.

* add : `TargetEncoder` preprocessing module added.

* [revert] : `TargetEncoder` implementation reverted.

* add : `QuantileTransformer` preprocessing model.

* prevent "object" dtype dictating

* prevent redundant reshaping

* `KBinsDiscretizer` added to `pymilo_param.py`

* add check prefix list function

* add `KBinsDiscretizer` test case

* fix dtype missing

* prevent index out of bound

* remove trailing whitespaces

* `kbins_discretizer` added to `test_preprocessings.py`

* remove trailing whitespaces

* add inner preprocessing module transportation (1 depth)

* `PowerTransformer` added

* add `PowerTransformer` test case

* `power_transformer` added to `test_preprocessings.py`

* `SplineTransformer` added

* `spline_transformer` added to `test_preprocessings.py`

* `SplineTransformer` test case added

* `BSpline` transporting added to `preprocessing_transporter.py`

* remove unused import

* add exception handling to`SplineTransformer` import

* add exception handling to`SplineTransformer` testcase

* add `TargetEncoder` to `pymilo_param.py`

* add `TargetEncoder` testcase

* TargetEncoder testcase added

* apply dtype dictation only if inner items doesn't have any

* enhance `target_encoder` to be comparable before after pymiloing

* remove unused variable + fixing the exception type

* `CHANGELOG.md` updated

* `CHANGELOG.md` updated

* add `numpy.nan` to NUMPY_TYPE_DICT

* add `NaN` type transportation

* `CHANGELOG.md` updated

* refactor function's core functionality and make it simpler

* make for iterator simpler & faster

* remove try, except and decrease the complexity

* fulfill preprocessing table

* `CHANGELOG.md` enhanced

* `autopep8.sh` applied

* fix bug

* `README.md` updated

* revert : changelog older log reverted.

* update : requirements updated.

* remove : extra line removed in `CHANGELOG.md`.

* update : last update tag in `SUPPORTED_MODELS.md` updated.

* change : `a` -> `list1` and `b` -> `list2`.

* add : `scipy` added to `dev-requirements.txt`.

* remove unused import has_named_parameter

---------

Co-authored-by: Sadra Sabouri <[email protected]>
Co-authored-by: AHReccese <[email protected]>

.gitignore

@@ -99,12 +99,5 @@ ENV/
 out
 gen
 
-/tests/exported_linear_models 
-/tests/exported_neural_networks
-/tests/exported_decision_trees
-/tests/exported_clusterings
-/tests/exported_naive_bayes
-/tests/exported_svms
-/tests/exported_neighbors
-/tests/exported_ensembles
 /.VSCodeCounter
+/tests/exported*

CHANGELOG.md

@@ -6,7 +6,32 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 
 ## [Unreleased]
 ### Added
+- `prefix_list` function in `utils.util.py`
+- `KBinsDiscretizer` preprocessing model
+- `PowerTransformer` preprocessing model
+- `SplineTransformer` preprocessing model
+- `TargetEncoder` preprocessing model
+- `QuantileTransformer` preprocessing model
+- `RobustScaler` preprocessing model
+- `PolynomialFeatures` preprocessing model
+- `OrdinalEncoder` preprocessing model
+- `Normalizer` preprocessing model
+- `MaxAbsScaler` preprocessing model
+- `MultiLabelBinarizer` preprocessing model
+- `KernelCenterer` preprocessing model
+- `FunctionTransformer` preprocessing model
+- `Binarizer` preprocessing model
+- Preprocessing models test runner
 ### Changed
+- `NaN` type in `pymilo_param`
+- `NaN` type transportation in `GeneralDataStructureTransporter` Transporter
+- `BSpline` Transportation in `PreprocessingTransporter` Transporter
+- one layer deeper transportation in `PreprocessingTransporter` Transporter
+- dictating outer ndarray dtype in `GeneralDataStructureTransporter` Transporter 
+- preprocessing params fulfilled in `pymilo_param`
+- `SUPPORTED_MODELS.md` updated
+- `README.md` updated
+- `serialize_possible_ml_model` in the Ensemble chain
 ## [0.8] - 2024-05-06
 ### Added
 - `StandardScaler` Transformer in `pymilo_param.py`

README.md

@@ -114,7 +114,7 @@ PyMilo is an open source Python package that provides a simple, efficient, and s
 | Nearest Neighbors ✅ | -  |  
 | Ensemble Models ✅ | - | 
 | Pipeline Model ✅ | - |
-| Preprocessing Models ⏳ | - |
+| Preprocessing Models ✅ | - |
 
 Details are available in [Supported Models](https://github.com/openscilab/pymilo/blob/main/SUPPORTED_MODELS.md).
 

SUPPORTED_MODELS.md

@@ -1,6 +1,6 @@
 # Supported Models
 
-**Last Update: 2024-04-24**
+**Last Update: 2024-05-30**
 
 
 <h2 id="scikit-learn">Scikit-Learn</h2> 
@@ -635,4 +635,75 @@
 		<td><b>StandardScaler</b></td>
         <td>>=0.8</td>
 	</tr>
+	<tr align="center">
+		<td>5</td>
+		<td><b>Binarizer</b></td>
+        <td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>6</td>
+		<td><b>FunctionTransformer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>7</td>
+		<td><b>KernelCenterer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>8</td>
+		<td><b>MultiLabelBinarizer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>9</td>
+		<td><b>MaxAbsScaler</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>10</td>
+		<td><b>Normalizer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>11</td>
+		<td><b>OrdinalEncoder</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>12</td>
+		<td><b>PolynomialFeatures</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>13</td>
+		<td><b>RobustScaler</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>14</td>
+		<td><b>QuantileTransformer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>15</td>
+		<td><b>KBinsDiscretizer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>16</td>
+		<td><b>PowerTransformer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>17</td>
+		<td><b>SplineTransformer</b></td>
+		<td>>=0.9</td>
+	</tr>
+	<tr align="center">
+		<td>18</td>
+		<td><b>TargetEncoder</b></td>
+		<td>>=0.9</td>
+	</tr>
+
 </table>

dev-requirements.txt

@@ -1,8 +1,9 @@
 numpy==1.26.4
 scikit-learn==1.5.0
+scipy>=0.19.1
 setuptools>=40.8.0
 vulture>=1.0
 bandit>=1.5.1
 pydocstyle>=3.0.0
 pytest>=4.3.1
-pytest-cov>=2.6.1
+pytest-cov>=2.6.1

otherfiles/meta.yaml

@@ -17,8 +17,10 @@ requirements:
         - setuptools
         - python >=3.6
     run:
-        - art >=1.8
         - python >=3.6
+        - numpy >=1.9.0
+        - scikit-learn >=0.22.2
+        - scipy >=0.19.1
 about:
     home: https://github.com/openscilab/pymilo
     license: MIT

pymilo/pymilo_param.py

@@ -67,6 +67,20 @@
 except BaseException:
     pass
 
+spline_transformer_support = False
+try:
+    from sklearn.preprocessing import SplineTransformer
+    spline_transformer_support = True
+except BaseException:
+    pass
+
+target_encoder_support = False
+try:
+    from sklearn.preprocessing import TargetEncoder
+    target_encoder_support = True
+except BaseException:
+    pass
+
 
 PYMILO_VERSION = "0.8"
 NOT_SUPPORTED = "NOT_SUPPORTED"
@@ -202,6 +216,20 @@
     "OneHotEncoder": preprocessing.OneHotEncoder,
     "LabelBinarizer": preprocessing.LabelBinarizer,
     "LabelEncoder": preprocessing.LabelEncoder,
+    "Binarizer": preprocessing.Binarizer,
+    "FunctionTransformer": preprocessing.FunctionTransformer,
+    "KernelCenterer": preprocessing.KernelCenterer,
+    "MultiLabelBinarizer": preprocessing.MultiLabelBinarizer,
+    "MaxAbsScaler": preprocessing.MaxAbsScaler,
+    "Normalizer": preprocessing.Normalizer,
+    "OrdinalEncoder": preprocessing.OrdinalEncoder,
+    "PolynomialFeatures": preprocessing.PolynomialFeatures,
+    "RobustScaler": preprocessing.RobustScaler,
+    "QuantileTransformer": preprocessing.QuantileTransformer,
+    "KBinsDiscretizer": preprocessing.KBinsDiscretizer,
+    "PowerTransformer": preprocessing.PowerTransformer,
+    "SplineTransformer": SplineTransformer if spline_transformer_support else NOT_SUPPORTED,
+    "TargetEncoder": TargetEncoder if target_encoder_support else NOT_SUPPORTED,
 }
 
 KEYS_NEED_PREPROCESSING_BEFORE_DESERIALIZATION = {
@@ -223,6 +251,7 @@
     "numpy.uint8": np.uint8,
     "numpy.uint64": np.uint64,
     "numpy.dtype": np.dtype,
+    "numpy.nan": np.nan,
 }
 
 EXPORTED_MODELS_PATH = {

pymilo/transporters/function_transporter.py

@@ -4,6 +4,7 @@
 from ..utils.util import import_function, check_str_in_iterable
 from .transporter import AbstractTransporter
 from types import FunctionType
+from numpy import ufunc
 
 array_function_dispatcher_support = False
 try:
@@ -30,12 +31,17 @@ def serialize(self, data, key, model_type):
         :type model_type: str
         :return: pymilo serialized output of data[key]
         """
-        if isinstance(
-                data[key],
-                FunctionType) or (
-                array_function_dispatcher_support and isinstance(
-                data[key],
-                _ArrayFunctionDispatcher)):
+        if isinstance(data[key], ufunc):
+            function = data[key]
+            data[key] = {
+                "function_name": function.__name__,
+                "function_module": "numpy",
+            }
+            return data[key]
+
+        elif isinstance(data[key], FunctionType) or (
+                array_function_dispatcher_support and
+                isinstance(data[key], _ArrayFunctionDispatcher)):
             function = data[key]
             data[key] = {
                 "function_name": function.__name__,

pymilo/transporters/general_data_structure_transporter.py

@@ -5,8 +5,8 @@
 
 from ..pymilo_param import NUMPY_TYPE_DICT
 
-from ..utils.util import get_homogeneous_type, all_same
-from ..utils.util import is_primitive, is_iterable, check_str_in_iterable
+from ..utils.util import get_homogeneous_type, all_same, prefix_list
+from ..utils.util import is_primitive, check_str_in_iterable
 
 from .transporter import AbstractTransporter
 
@@ -93,7 +93,13 @@ def serialize(self, data, key, model_type):
         :type model_type: str
         :return: pymilo serialized output of data[key]
         """
-        if isinstance(data[key], type):
+        if not (isinstance(data[key], object) or isinstance(data[key], str)):
+            if np.isnan(data[key]):  # throws exception on object & str types
+                data[key] = {
+                    "np-type": "numpy.nan",
+                    "value": "NaN"
+                }
+        elif isinstance(data[key], type):
             raw_type = str(data[key])
             raw_type = "numpy" + str(raw_type).split("numpy")[-1][:-2]
             if raw_type in NUMPY_TYPE_DICT.keys():
@@ -281,6 +287,8 @@ def get_deserialized_regular_primary_types(self, content):
         if "np-type" in content:
             if content["np-type"] == "numpy.dtype":
                 return NUMPY_TYPE_DICT[content["np-type"]](NUMPY_TYPE_DICT[content['value']])
+            if content["np-type"] == "numpy.nan":
+                return NUMPY_TYPE_DICT[content["np-type"]]
             return NUMPY_TYPE_DICT[content["np-type"]](content['value'])
 
     def is_numpy_primary_type(self, content):
@@ -450,4 +458,11 @@ def deep_deserialize_ndarray(self, deserialized_ndarray):
                         new_list.append(item)
                 else:
                     new_list.append(item)
-        return np.asarray(new_list, dtype=dtype).reshape(shape)
+
+        pre_result = np.asarray(new_list, dtype=dtype)
+        if dtype == "object" and hasattr(new_list[0], "dtype"):
+            # check if inner items have specific dtype.
+            pre_result = np.asarray(new_list)
+        if not prefix_list(list(pre_result.shape), shape):
+            return pre_result.reshape(shape)
+        return pre_result