Skip to content

octopus.study

Study module.

OctoClassification

Bases: OctoStudy

Classification study (binary and multiclass).

Source code in octopus/study/core.py 
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
@define
class OctoClassification(OctoStudy):
    """Classification study (binary and multiclass)."""

    target_col: str = field(kw_only=True, validator=validators.instance_of(str))
    """The target column to predict."""

    ml_type: MLType | None = field(  # type: ignore[assignment]
        default=None,
        kw_only=True,
        validator=validators.optional(validators.in_([MLType.BINARY, MLType.MULTICLASS])),
    )
    """The type of machine learning model. Can be set explicitly or auto-detected from data (binary vs multiclass)."""

    target_metric: str = field(
        default="AUCROC",
        validator=validators.in_(Metrics.get_by_type(MLType.BINARY, MLType.MULTICLASS)),
    )
    """The primary metric used for model evaluation. Defaults to AUCROC."""

    positive_class: int | None = field(default=None, validator=validators.optional(validators.instance_of(int)))
    """The positive class label for binary classification. Defaults to None. Not used for multiclass."""

    def _resolve_ml_config(self, data: pd.DataFrame) -> tuple[MLType, int | None]:
        if self.target_col not in data.columns:
            raise ValueError(f"Target column '{self.target_col}' not found in input data.")
        ml_type = self.ml_type
        positive_class = self.positive_class
        if not ml_type:
            unique_values = data[self.target_col].dropna().unique()
            if len(unique_values) > 2:
                ml_type, positive_class = MLType.MULTICLASS, None
            else:
                ml_type, positive_class = MLType.BINARY, 1
        if ml_type == MLType.BINARY and positive_class is None:
            raise ValueError("For binary classification, `positive_class` must be specified.")
        return ml_type, positive_class

    def _validate_data(self, data: pd.DataFrame, ml_type: MLType, positive_class: int | None) -> None:
        if self.target_col not in data.columns:
            raise ValueError(f"Target column '{self.target_col}' not found in input data.")
        super()._validate_data(data, ml_type, positive_class)

    @property
    def target_assignments(self) -> dict[str, str]:
        """Get target assignments dict."""
        return {"default": self.target_col}

ml_type = field(default=None, kw_only=True, validator=(validators.optional(validators.in_([MLType.BINARY, MLType.MULTICLASS])))) class-attribute instance-attribute

The type of machine learning model. Can be set explicitly or auto-detected from data (binary vs multiclass).

positive_class = field(default=None, validator=(validators.optional(validators.instance_of(int)))) class-attribute instance-attribute

The positive class label for binary classification. Defaults to None. Not used for multiclass.

target_assignments property

Get target assignments dict.

target_col = field(kw_only=True, validator=(validators.instance_of(str))) class-attribute instance-attribute

The target column to predict.

target_metric = field(default='AUCROC', validator=(validators.in_(Metrics.get_by_type(MLType.BINARY, MLType.MULTICLASS)))) class-attribute instance-attribute

The primary metric used for model evaluation. Defaults to AUCROC.

OctoRegression

Bases: OctoStudy

Regression study.

Source code in octopus/study/core.py 
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
@define
class OctoRegression(OctoStudy):
    """Regression study."""

    target_col: str = field(kw_only=True, validator=validators.instance_of(str))
    """The target column to predict."""

    ml_type: MLType = field(default=MLType.REGRESSION, init=False)
    """The type of machine learning model. Automatically set to regression."""

    target_metric: str = field(
        default="RMSE",
        validator=validators.in_(Metrics.get_by_type(MLType.REGRESSION)),
    )
    """The primary metric used for model evaluation. Defaults to RMSE."""

    @property
    def target_assignments(self) -> dict[str, str]:
        """Get target assignments dict."""
        return {"default": self.target_col}

ml_type = field(default=(MLType.REGRESSION), init=False) class-attribute instance-attribute

The type of machine learning model. Automatically set to regression.

target_assignments property

Get target assignments dict.

target_col = field(kw_only=True, validator=(validators.instance_of(str))) class-attribute instance-attribute

The target column to predict.

target_metric = field(default='RMSE', validator=(validators.in_(Metrics.get_by_type(MLType.REGRESSION)))) class-attribute instance-attribute

The primary metric used for model evaluation. Defaults to RMSE.

OctoStudy

Bases: ABC

Abstract base class for all Octopus studies.

Source code in octopus/study/core.py 
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 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
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
@define
class OctoStudy(ABC):
    """Abstract base class for all Octopus studies."""

    feature_cols: list[str] = field(validator=[validators.instance_of(list)])
    """List of all feature columns in the dataset."""

    sample_id_col: str = field(validator=validators.instance_of(str))
    """Identifier for sample instances."""

    name: str = field(default="Octopus", validator=[validators.instance_of(str)])
    """The name of the study. Defaults to 'Octopus'."""

    row_id_col: str | None = field(
        default=Factory(lambda: None),
        validator=validators.optional(validators.instance_of(str)),
    )
    """Unique row identifier."""

    stratification_col: str | None = field(
        default=Factory(lambda: None),
        validator=validators.optional(validators.instance_of(str)),
    )
    """Column used for stratification during data splitting."""

    n_folds_outer: int = field(default=5 if not _RUNNING_IN_TESTSUITE else 2, validator=[validators.instance_of(int)])
    """The number of outer folds for cross-validation. Defaults to 5."""

    datasplit_seed_outer: int = field(default=0, validator=[validators.instance_of(int)])
    """The seed used for data splitting in outer cross-validation. Defaults to 0."""

    ignore_data_health_warning: bool = field(default=Factory(lambda: False), validator=[validators.instance_of(bool)])
    """Ignore data health checks warning and run machine learning workflow."""

    outer_parallelization: bool = field(default=Factory(lambda: True), validator=[validators.instance_of(bool)])
    """Indicates whether outer parallelization is enabled. Defaults to True."""

    run_single_outersplit_num: int = field(default=Factory(lambda: -1), validator=[validators.instance_of(int)])
    """Select a single outersplit to execute. Defaults to -1 to run all outersplits"""

    workflow: Sequence[Task] = field(
        default=Factory(lambda: [Octo(task_id=0)]),
        validator=[validators.instance_of(list), validate_workflow],
    )
    """A list of tasks that defines the processing workflow. Each item in the list is an instance of `Task`."""

    path: UPath = field(default=UPath("./studies/"), converter=lambda x: UPath(x))
    """The path where study outputs are saved. Defaults to "./studies/"."""

    ml_type: MLType = field(init=False)
    """The type of machine learning model. Set automatically by subclass."""

    # Time of last fit() call (internal state)
    _fit_timestamp: str | None = field(default=None, init=False)

    @property
    @abstractmethod
    def target_metric(self) -> str:
        """Get target metric. Must be implemented in subclasses."""
        ...

    @property
    @abstractmethod
    def target_assignments(self) -> dict[str, str]:
        """Get target assignments dict. Must be implemented in subclasses."""
        ...

    @property
    def output_path(self) -> UPath:
        """Full output path for this study (path/name-timestamp)."""
        if self._fit_timestamp is None:
            raise RuntimeError("output_path is not available until fit() has been called.")
        fit_dt = datetime.datetime.fromisoformat(self._fit_timestamp)
        folder_name = f"{self.name}-{fit_dt.strftime('%Y%m%d_%H%M%S')}"
        return self.path / folder_name

    @property
    def log_dir(self) -> UPath:
        """Directory where logs are stored."""
        return self.output_path

    def _resolve_ml_config(self, data: pd.DataFrame) -> tuple[MLType, int | None]:
        """Resolve ml_type and positive_class. Subclasses override for auto-detection."""
        return self.ml_type, getattr(self, "positive_class", None)

    def _validate_data(self, data: pd.DataFrame, ml_type: MLType, positive_class: int | None) -> None:
        """Validate the input data."""
        validator = OctoDataValidator(
            data=data,
            feature_cols=self.feature_cols,
            target_col=self.target_col if hasattr(self, "target_col") else None,
            duration_col=self.duration_col if hasattr(self, "duration_col") else None,
            event_col=self.event_col if hasattr(self, "event_col") else None,
            sample_id_col=self.sample_id_col,
            row_id_col=self.row_id_col,
            stratification_col=self.stratification_col,
            ml_type=ml_type,
            positive_class=positive_class,
        )
        validator.validate()

    def _initialize_study_directory(self) -> None:
        """Initialize study directory."""
        if self.output_path.exists():
            raise FileExistsError(f"Study output folder already exists: {self.output_path}")
        self.output_path.mkdir(parents=True, exist_ok=False)
        self.log_dir.mkdir(parents=True, exist_ok=True)
        set_logger_filename(log_file=self.log_dir / "study.log")

    def _initialize_study_outputs(
        self,
        data: pd.DataFrame,
        prepared: PreparedData,
        ml_type: MLType,
        positive_class: int | None,
    ) -> None:
        """Initialize study saving config and data into study directory."""

        def serialize_value(value):
            """Convert a value to JSON-serializable format."""
            if hasattr(value, "value"):
                return value.value
            elif isinstance(value, UPath):
                return str(value)
            elif has(type(value)):
                # Convert to dict using asdict
                result = asdict(value, value_serializer=lambda _, __, v: serialize_value(v))

                # Add ClassVar 'module' field if it exists (for workflow tasks)
                if hasattr(value, "module"):
                    result["module"] = value.module

                return result
            elif isinstance(value, list):
                return [serialize_value(item) for item in value]
            elif isinstance(value, dict):
                return {k: serialize_value(v) for k, v in value.items()}
            return value

        config = {}
        # Use fields from the actual instance class (including subclass fields)
        for attr in fields(type(self)):
            value = getattr(self, attr.name)
            config[attr.name] = serialize_value(value)

        # Override with resolved values (e.g. auto-detected ml_type for OctoClassification)
        config["ml_type"] = ml_type.value
        if "positive_class" in config:
            config["positive_class"] = positive_class
        config["prepared"] = {
            "feature_cols": prepared.feature_cols,
            "row_id": prepared.row_id_col,
            "target_assignments": self.target_assignments,
        }

        config_path = self.output_path / "study_config.json"
        with config_path.open("w") as f:
            json.dump(config, f, indent=2)

        # Write study metadata (version, platform, timestamp)
        assert self._fit_timestamp is not None  # Set at start of fit()
        study_meta = {
            "octopus_version": get_version(),
            "package_name": get_package_name(),
            "python_version": platform.python_version(),
            "created_at": self._fit_timestamp,
        }
        meta_path = self.output_path / "study_meta.json"
        with meta_path.open("w") as f:
            json.dump(study_meta, f, indent=2)

        data_path = self.output_path / "data_raw.parquet"
        parquet_save(data, data_path, index=False)
        prepared_data_path = self.output_path / "data_prepared.parquet"
        parquet_save(prepared.data, prepared_data_path, index=False)

    def _prepare_data(self, data: pd.DataFrame) -> PreparedData:
        """Prepare the data for training."""
        preparator = OctoDataPreparator(
            data=data,
            feature_cols=self.feature_cols,
            sample_id_col=self.sample_id_col,
            row_id_col=self.row_id_col,
            target_col=self.target_col if hasattr(self, "target_col") else None,
            stratification_col=self.stratification_col,
            duration_col=self.duration_col if hasattr(self, "duration_col") else None,
            event_col=self.event_col if hasattr(self, "event_col") else None,
        )
        return preparator.prepare()

    def _create_datasplits(self, prepared: PreparedData) -> OuterSplits:
        """Create datasplits for outer cross-validation."""
        relevant_cols = list(prepared.feature_cols) + [
            c
            for c in (
                self.target_col if hasattr(self, "target_col") else None,
                self.duration_col if hasattr(self, "duration_col") else None,
                self.event_col if hasattr(self, "event_col") else None,
                self.sample_id_col,
                prepared.row_id_col,
                self.stratification_col,
            )
            if c is not None
        ]

        relevant_cols += [DATASPLIT_COL]

        relevant_cols = list(dict.fromkeys(relevant_cols))
        data_clean = prepared.data[relevant_cols]

        outersplits = DataSplit(
            dataset=data_clean,
            seeds=[self.datasplit_seed_outer],
            num_folds=self.n_folds_outer,
            stratification_col=self.stratification_col,
        ).get_outer_splits()

        return outersplits

    def _run_health_check(self, prepared: PreparedData, config: HealthCheckConfig | None) -> None:
        """Run data health check, save results, and check for issues."""
        checker = OctoDataHealthChecker(
            data=prepared.data,
            feature_cols=prepared.feature_cols,
            target_col=self.target_col if hasattr(self, "target_col") else None,
            duration_col=self.duration_col if hasattr(self, "duration_col") else None,
            event_col=self.event_col if hasattr(self, "event_col") else None,
            row_id_col=prepared.row_id_col,
            sample_id_col=self.sample_id_col,
            stratification_col=self.stratification_col,
            config=config or HealthCheckConfig(),
        )
        report = checker.generate_report()
        report_path = self.output_path / "health_check_report.csv"
        csv_save(report, report_path, index=False)

        if report.empty:
            return

        has_critical = False
        has_warning = False

        if "Severity" in report.columns:
            has_critical = (report["Severity"] == "Critical").any()
            has_warning = (report["Severity"] == "Warning").any()

        if has_critical:
            raise ValueError(f"Critical data issues detected. Please check: {report_path}")

        if has_warning and not self.ignore_data_health_warning:
            raise ValueError(
                f"Data issues detected. Please check: {report_path}\nTo proceed despite warnings, set `ignore_data_health_warning=True`."
            )

    def _flush_logger(self):
        """Flush and close the file log handler.

        For remote filesystems (e.g. S3 via fsspec), log data is buffered
        locally and only uploaded when the stream is closed. This method
        ensures all buffered log entries are written before the study
        finishes.
        """
        # set_logger_filename(None) finds FSSpecFileHandlers, closes them
        # (which flushes + uploads for remote FS), and removes them.
        set_logger_filename(log_file=None)

    def _create_study_context(
        self, prepared: PreparedData, ml_type: MLType, positive_class: int | None
    ) -> StudyContext:
        """Create a frozen StudyContext from the current study state."""
        return StudyContext(
            ml_type=ml_type,
            target_metric=self.target_metric,
            target_assignments=self.target_assignments,
            positive_class=positive_class,
            stratification_col=self.stratification_col,
            sample_id_col=self.sample_id_col,
            feature_cols=prepared.feature_cols,
            row_id_col=prepared.row_id_col,
            output_path=self.output_path,
            log_dir=self.log_dir,
        )

    def fit(
        self,
        data: pd.DataFrame,
        health_check_config: HealthCheckConfig | None = None,
    ) -> None:
        """Fit study to data.

        Args:
            data: DataFrame containing the dataset.
            health_check_config: Optional configuration for health check thresholds.
        """
        if self._fit_timestamp is not None:
            raise RuntimeError("fit() can only be called once per study instance.")

        # Generate single timestamp for this fit() call
        self._fit_timestamp = datetime.datetime.now(UTC).isoformat()

        self._initialize_study_directory()
        ml_type, positive_class = self._resolve_ml_config(data)
        self._validate_data(data, ml_type, positive_class)
        prepared = self._prepare_data(data)
        self._initialize_study_outputs(data, prepared, ml_type, positive_class)
        self._run_health_check(prepared, health_check_config)

        outersplit_data = self._create_datasplits(prepared)
        study_context = self._create_study_context(prepared, ml_type, positive_class)
        manager = OctoManager(
            outersplit_data=outersplit_data,
            study_context=study_context,
            workflow=self.workflow,
            outer_parallelization=self.outer_parallelization,
            run_single_outersplit_num=self.run_single_outersplit_num,
        )
        manager.run_outersplits()
        logger.info("Study completed. Results saved to: %s", self.output_path)
        self._flush_logger()

datasplit_seed_outer = field(default=0, validator=[validators.instance_of(int)]) class-attribute instance-attribute

The seed used for data splitting in outer cross-validation. Defaults to 0.

feature_cols = field(validator=[validators.instance_of(list)]) class-attribute instance-attribute

List of all feature columns in the dataset.

ignore_data_health_warning = field(default=(Factory(lambda: False)), validator=[validators.instance_of(bool)]) class-attribute instance-attribute

Ignore data health checks warning and run machine learning workflow.

log_dir property

Directory where logs are stored.

ml_type = field(init=False) class-attribute instance-attribute

The type of machine learning model. Set automatically by subclass.

n_folds_outer = field(default=(5 if not _RUNNING_IN_TESTSUITE else 2), validator=[validators.instance_of(int)]) class-attribute instance-attribute

The number of outer folds for cross-validation. Defaults to 5.

name = field(default='Octopus', validator=[validators.instance_of(str)]) class-attribute instance-attribute

The name of the study. Defaults to 'Octopus'.

outer_parallelization = field(default=(Factory(lambda: True)), validator=[validators.instance_of(bool)]) class-attribute instance-attribute

Indicates whether outer parallelization is enabled. Defaults to True.

output_path property

Full output path for this study (path/name-timestamp).

path = field(default=(UPath('./studies/')), converter=(lambda x: UPath(x))) class-attribute instance-attribute

The path where study outputs are saved. Defaults to "./studies/".

row_id_col = field(default=(Factory(lambda: None)), validator=(validators.optional(validators.instance_of(str)))) class-attribute instance-attribute

Unique row identifier.

run_single_outersplit_num = field(default=(Factory(lambda: -1)), validator=[validators.instance_of(int)]) class-attribute instance-attribute

Select a single outersplit to execute. Defaults to -1 to run all outersplits

sample_id_col = field(validator=(validators.instance_of(str))) class-attribute instance-attribute

Identifier for sample instances.

stratification_col = field(default=(Factory(lambda: None)), validator=(validators.optional(validators.instance_of(str)))) class-attribute instance-attribute

Column used for stratification during data splitting.

target_assignments abstractmethod property

Get target assignments dict. Must be implemented in subclasses.

target_metric abstractmethod property

Get target metric. Must be implemented in subclasses.

workflow = field(default=(Factory(lambda: [Octo(task_id=0)])), validator=[validators.instance_of(list), validate_workflow]) class-attribute instance-attribute

A list of tasks that defines the processing workflow. Each item in the list is an instance of Task.

fit(data, health_check_config=None)

Fit study to data.

Parameters:

Name Type Description Default
data DataFrame

DataFrame containing the dataset.

 required
health_check_config HealthCheckConfig | None

Optional configuration for health check thresholds.

 None
Source code in octopus/study/core.py 
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
def fit(
    self,
    data: pd.DataFrame,
    health_check_config: HealthCheckConfig | None = None,
) -> None:
    """Fit study to data.

    Args:
        data: DataFrame containing the dataset.
        health_check_config: Optional configuration for health check thresholds.
    """
    if self._fit_timestamp is not None:
        raise RuntimeError("fit() can only be called once per study instance.")

    # Generate single timestamp for this fit() call
    self._fit_timestamp = datetime.datetime.now(UTC).isoformat()

    self._initialize_study_directory()
    ml_type, positive_class = self._resolve_ml_config(data)
    self._validate_data(data, ml_type, positive_class)
    prepared = self._prepare_data(data)
    self._initialize_study_outputs(data, prepared, ml_type, positive_class)
    self._run_health_check(prepared, health_check_config)

    outersplit_data = self._create_datasplits(prepared)
    study_context = self._create_study_context(prepared, ml_type, positive_class)
    manager = OctoManager(
        outersplit_data=outersplit_data,
        study_context=study_context,
        workflow=self.workflow,
        outer_parallelization=self.outer_parallelization,
        run_single_outersplit_num=self.run_single_outersplit_num,
    )
    manager.run_outersplits()
    logger.info("Study completed. Results saved to: %s", self.output_path)
    self._flush_logger()

OctoTimeToEvent

Bases: OctoStudy

Time-to-event study.

Source code in octopus/study/core.py 
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
@define
class OctoTimeToEvent(OctoStudy):
    """Time-to-event study."""

    duration_col: str = field(kw_only=True, validator=validators.instance_of(str))
    """Column containing time until event or censoring."""

    event_col: str = field(kw_only=True, validator=validators.instance_of(str))
    """Column containing event indicator (0=censored, 1=event)."""

    ml_type: MLType = field(default=MLType.TIMETOEVENT, init=False)
    """The type of machine learning model. Automatically set to time-to-event."""

    target_metric: str = field(
        default="CI",
        validator=validators.in_(Metrics.get_by_type(MLType.TIMETOEVENT)),
    )
    """The primary metric used for model evaluation. Defaults to CI (Concordance Index)."""

    @property
    def target_assignments(self) -> dict[str, str]:
        """Get target assignments dict."""
        return {"duration": self.duration_col, "event": self.event_col}

duration_col = field(kw_only=True, validator=(validators.instance_of(str))) class-attribute instance-attribute

Column containing time until event or censoring.

event_col = field(kw_only=True, validator=(validators.instance_of(str))) class-attribute instance-attribute

Column containing event indicator (0=censored, 1=event).

ml_type = field(default=(MLType.TIMETOEVENT), init=False) class-attribute instance-attribute

The type of machine learning model. Automatically set to time-to-event.

target_assignments property

Get target assignments dict.

target_metric = field(default='CI', validator=(validators.in_(Metrics.get_by_type(MLType.TIMETOEVENT)))) class-attribute instance-attribute

The primary metric used for model evaluation. Defaults to CI (Concordance Index).