Skip to content

API: preprocessing.split

skyulf.preprocessing.split

DataSplitter

Splits a DataFrame into Train, Test, and optionally Validation sets.

Source code in skyulf-core\skyulf\preprocessing\split.py
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
class DataSplitter:
    """
    Splits a DataFrame into Train, Test, and optionally Validation sets.
    """

    def __init__(
        self,
        test_size: float = 0.2,
        validation_size: float = 0.0,
        random_state: int = 42,
        shuffle: bool = True,
        stratify_col: Optional[str] = None,
    ):
        self.test_size = test_size
        self.validation_size = validation_size
        self.random_state = random_state
        self.shuffle = shuffle
        self.stratify_col = stratify_col

    def split_xy(self, X: Union[pd.DataFrame, SkyulfDataFrame], y: Union[pd.Series, Any]) -> SplitDataset:
        """
        Splits X and y arrays.
        """
        engine = get_engine(X)
        is_polars = engine.name == "polars"

        if is_polars:
            # Convert to Pandas to preserve schema/metadata during split
            X_pd = X.to_pandas()
            y_pd = y.to_pandas() if y is not None else None
        else:
            X_pd = X
            y_pd = y

        stratify = y_pd if self.stratify_col else None  # If stratify is requested, use y

        if stratify is not None:
            # Check value counts
            class_counts = y_pd.value_counts()
            min_count = class_counts.min()

            if min_count < 2:
                logger.warning(
                    f"Stratified split requested but the least populated class has only {min_count} "
                    "member(s). Stratification will be disabled."
                )
                stratify = None

        # First split: Train+Val vs Test
        X_train_val, X_test, y_train_val, y_test = train_test_split(
            X_pd,
            y_pd,
            test_size=self.test_size,
            random_state=self.random_state,
            shuffle=self.shuffle,
            stratify=stratify,
        )

        validation = None
        if self.validation_size > 0:
            relative_val_size = self.validation_size / (1 - self.test_size)
            stratify_val = y_train_val if self.stratify_col else None

            if stratify_val is not None:
                class_counts_val = y_train_val.value_counts()
                min_count_val = class_counts_val.min()

                if min_count_val < 2:
                    logger.warning(
                        "Stratified validation split requested but the least populated class has only "
                        f"{min_count_val} member(s). Stratification will be disabled for validation split."
                    )
                    stratify_val = None

            X_train, X_val, y_train, y_val = train_test_split(
                X_train_val,
                y_train_val,
                test_size=relative_val_size,
                random_state=self.random_state,
                shuffle=self.shuffle,
                stratify=stratify_val,
            )
            validation = (X_val, y_val)
        else:
            X_train, y_train = X_train_val, y_train_val

        # Convert back to Polars if needed
        if is_polars:
            import polars as pl

            def to_pl(df_or_series):
                if df_or_series is None: return None
                if isinstance(df_or_series, pd.DataFrame): return pl.from_pandas(df_or_series)
                if isinstance(df_or_series, pd.Series): return pl.from_pandas(df_or_series)
                return df_or_series

            X_train = to_pl(X_train)
            y_train = to_pl(y_train)
            X_test = to_pl(X_test)
            y_test = to_pl(y_test)

            if validation:
                validation = (to_pl(validation[0]), to_pl(validation[1]))

        return SplitDataset(
            train=(X_train, y_train), test=(X_test, y_test), validation=validation
        )

    def split(self, df: Union[pd.DataFrame, SkyulfDataFrame]) -> SplitDataset:
        """
        Splits a DataFrame.
        """
        engine = get_engine(df)
        is_polars = engine.name == "polars"

        if is_polars:
            # Convert to Pandas to preserve schema/metadata during split
            df_pd = df.to_pandas()
        else:
            df_pd = df

        stratify = None
        if self.stratify_col and self.stratify_col in df_pd.columns:
            stratify = df_pd[self.stratify_col]
            class_counts = stratify.value_counts()
            if class_counts.min() < 2:
                logger.warning(
                    f"Stratified split requested but the least populated class has only {class_counts.min()} "
                    "member(s). Stratification will be disabled."
                )
                stratify = None

        train_val, test = train_test_split(
            df_pd,
            test_size=self.test_size,
            random_state=self.random_state,
            shuffle=self.shuffle,
            stratify=stratify,
        )

        validation = None
        if self.validation_size > 0:
            relative_val_size = self.validation_size / (1 - self.test_size)

            stratify_val = None
            if self.stratify_col and self.stratify_col in train_val.columns:
                stratify_val = train_val[self.stratify_col]
                class_counts_val = stratify_val.value_counts()
                if class_counts_val.min() < 2:
                    logger.warning(
                        "Stratified validation split requested but the least populated class has only "
                        f"{class_counts_val.min()} member(s). Stratification will be disabled for validation split."
                    )
                    stratify_val = None

            train, val = train_test_split(
                train_val,
                test_size=relative_val_size,
                random_state=self.random_state,
                shuffle=self.shuffle,
                stratify=stratify_val,
            )
            validation = val
        else:
            train = train_val

        # Convert back to Polars if needed
        if is_polars:
            import polars as pl
            train = pl.from_pandas(train)
            test = pl.from_pandas(test)
            if validation is not None:
                validation = pl.from_pandas(validation)

        return SplitDataset(train=train, test=test, validation=validation)

split(df)

Splits a DataFrame.

Source code in skyulf-core\skyulf\preprocessing\split.py
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
def split(self, df: Union[pd.DataFrame, SkyulfDataFrame]) -> SplitDataset:
    """
    Splits a DataFrame.
    """
    engine = get_engine(df)
    is_polars = engine.name == "polars"

    if is_polars:
        # Convert to Pandas to preserve schema/metadata during split
        df_pd = df.to_pandas()
    else:
        df_pd = df

    stratify = None
    if self.stratify_col and self.stratify_col in df_pd.columns:
        stratify = df_pd[self.stratify_col]
        class_counts = stratify.value_counts()
        if class_counts.min() < 2:
            logger.warning(
                f"Stratified split requested but the least populated class has only {class_counts.min()} "
                "member(s). Stratification will be disabled."
            )
            stratify = None

    train_val, test = train_test_split(
        df_pd,
        test_size=self.test_size,
        random_state=self.random_state,
        shuffle=self.shuffle,
        stratify=stratify,
    )

    validation = None
    if self.validation_size > 0:
        relative_val_size = self.validation_size / (1 - self.test_size)

        stratify_val = None
        if self.stratify_col and self.stratify_col in train_val.columns:
            stratify_val = train_val[self.stratify_col]
            class_counts_val = stratify_val.value_counts()
            if class_counts_val.min() < 2:
                logger.warning(
                    "Stratified validation split requested but the least populated class has only "
                    f"{class_counts_val.min()} member(s). Stratification will be disabled for validation split."
                )
                stratify_val = None

        train, val = train_test_split(
            train_val,
            test_size=relative_val_size,
            random_state=self.random_state,
            shuffle=self.shuffle,
            stratify=stratify_val,
        )
        validation = val
    else:
        train = train_val

    # Convert back to Polars if needed
    if is_polars:
        import polars as pl
        train = pl.from_pandas(train)
        test = pl.from_pandas(test)
        if validation is not None:
            validation = pl.from_pandas(validation)

    return SplitDataset(train=train, test=test, validation=validation)

split_xy(X, y)

Splits X and y arrays.

Source code in skyulf-core\skyulf\preprocessing\split.py
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
def split_xy(self, X: Union[pd.DataFrame, SkyulfDataFrame], y: Union[pd.Series, Any]) -> SplitDataset:
    """
    Splits X and y arrays.
    """
    engine = get_engine(X)
    is_polars = engine.name == "polars"

    if is_polars:
        # Convert to Pandas to preserve schema/metadata during split
        X_pd = X.to_pandas()
        y_pd = y.to_pandas() if y is not None else None
    else:
        X_pd = X
        y_pd = y

    stratify = y_pd if self.stratify_col else None  # If stratify is requested, use y

    if stratify is not None:
        # Check value counts
        class_counts = y_pd.value_counts()
        min_count = class_counts.min()

        if min_count < 2:
            logger.warning(
                f"Stratified split requested but the least populated class has only {min_count} "
                "member(s). Stratification will be disabled."
            )
            stratify = None

    # First split: Train+Val vs Test
    X_train_val, X_test, y_train_val, y_test = train_test_split(
        X_pd,
        y_pd,
        test_size=self.test_size,
        random_state=self.random_state,
        shuffle=self.shuffle,
        stratify=stratify,
    )

    validation = None
    if self.validation_size > 0:
        relative_val_size = self.validation_size / (1 - self.test_size)
        stratify_val = y_train_val if self.stratify_col else None

        if stratify_val is not None:
            class_counts_val = y_train_val.value_counts()
            min_count_val = class_counts_val.min()

            if min_count_val < 2:
                logger.warning(
                    "Stratified validation split requested but the least populated class has only "
                    f"{min_count_val} member(s). Stratification will be disabled for validation split."
                )
                stratify_val = None

        X_train, X_val, y_train, y_val = train_test_split(
            X_train_val,
            y_train_val,
            test_size=relative_val_size,
            random_state=self.random_state,
            shuffle=self.shuffle,
            stratify=stratify_val,
        )
        validation = (X_val, y_val)
    else:
        X_train, y_train = X_train_val, y_train_val

    # Convert back to Polars if needed
    if is_polars:
        import polars as pl

        def to_pl(df_or_series):
            if df_or_series is None: return None
            if isinstance(df_or_series, pd.DataFrame): return pl.from_pandas(df_or_series)
            if isinstance(df_or_series, pd.Series): return pl.from_pandas(df_or_series)
            return df_or_series

        X_train = to_pl(X_train)
        y_train = to_pl(y_train)
        X_test = to_pl(X_test)
        y_test = to_pl(y_test)

        if validation:
            validation = (to_pl(validation[0]), to_pl(validation[1]))

    return SplitDataset(
        train=(X_train, y_train), test=(X_test, y_test), validation=validation
    )