MODULE REFERENCE

observations

anomx.data.views.observations

Canonical observation container shared by every anomx pipeline.

An ObservationSet is a collection of observations plus optional structural information. Independent samples, sequences, time series, panel data, and graph-backed observations are all expressed through the same container; algorithms consume standardized views instead of raw storage formats.

classView source ↗

ObservationSet

ObservationSet(observations: pd.DataFrame, feature_columns: list[str] = <factory: list>, target_columns: list[str] = <factory: list>, entity_column: str | None = None, sequence_column: str | None = None, timestamp_column: str | None = None, position_column: str | None = None, relations: RelationSet | None = None, hierarchy: Hierarchy | None = None, metadata: dict[str, Any] = <factory: dict>)

A table of observations together with optional structural overlays.

Signature parameters
Parameter	Type	Default
`observations`	`pd.DataFrame`	`required`
`feature_columns`	`list[str]`	`<factory: list>`
`target_columns`	`list[str]`	`<factory: list>`
`entity_column`	`str \| None`	`None`
`sequence_column`	`str \| None`	`None`
`timestamp_column`	`str \| None`	`None`
`position_column`	`str \| None`	`None`
`relations`	`RelationSet \| None`	`None`
`hierarchy`	`Hierarchy \| None`	`None`
`metadata`	`dict[str, Any]`	`<factory: dict>`

Members

observationsattribute

# ObservationSet.observations Source ↗

observations: pd.DataFrame

feature_columnsattribute

# ObservationSet.feature_columns Source ↗

feature_columns: list[str] = field(default_factory=list)

target_columnsattribute

# ObservationSet.target_columns Source ↗

target_columns: list[str] = field(default_factory=list)

entity_columnattribute

# ObservationSet.entity_column Source ↗

entity_column: str | None = None

sequence_columnattribute

# ObservationSet.sequence_column Source ↗

sequence_column: str | None = None

timestamp_columnattribute

# ObservationSet.timestamp_column Source ↗

timestamp_column: str | None = None

position_columnattribute

# ObservationSet.position_column Source ↗

position_column: str | None = None

relationsattribute

# ObservationSet.relations Source ↗

relations: RelationSet | None = None

hierarchyattribute

# ObservationSet.hierarchy Source ↗

hierarchy: Hierarchy | None = None

metadataattribute

# ObservationSet.metadata Source ↗

metadata: dict[str, Any] = field(default_factory=dict)

observation_countproperty

# ObservationSet.observation_count Source ↗

observation_count: int

Returns int

characteristicsmethod

# ObservationSet.characteristics Source ↗

characteristics() -> DataCharacteristics

Derive the orthogonal structural properties of this set.

Returns DataCharacteristics

as_tabularmethod

# ObservationSet.as_tabular Source ↗

as_tabular() -> TabularView

Materialize an unordered n_samples × n_features matrix.

Returns TabularView

as_sequencesmethod

# ObservationSet.as_sequences Source ↗

as_sequences(*, group_by: str | None = None) -> SequenceView

Materialize per-sequence ordered frames.

Signature parameters
Parameter	Type	Default
`group_by`keyword only	`str \| None`	`None`

Returns SequenceView

as_windowsmethod

# ObservationSet.as_windows Source ↗

as_windows(*, window_size: int, stride: int = 1, group_by: str | None = None) -> WindowView

Materialize fixed-size sliding windows over ordered observations.

Signature parameters
Parameter	Type	Default
`window_size`keyword only	`int`	`required`
`stride`keyword only	`int`	`1`
`group_by`keyword only	`str \| None`	`None`

Returns WindowView

as_timeseriesmethod

# ObservationSet.as_timeseries Source ↗

as_timeseries() -> TimeSeriesView

Materialize a time-indexed frame for temporal models.

Returns TimeSeriesView

as_graphmethod

# ObservationSet.as_graph Source ↗

as_graph() -> GraphView

Materialize node features with the relational overlay.

Returns GraphView

attributeView source ↗

REGULAR_SAMPLING_RELATIVE_TOLERANCE

REGULAR_SAMPLING_RELATIVE_TOLERANCE = 0.01