Prediction

Classes to perform multidiversity index prediction.

BaseDataManager

Base class for data management. Contain layers to extract features needed for computation.

compute() abstractmethod

Compute sample from data layer and output a set of variables as Sample.

Source code in niva/core/prediction/data_manager.py

@abstractmethod
def compute(self) -> Sample:
    """Compute sample from data layer and output a set of variables as Sample."""
    pass

filter_iacs(layer)

Filter IACS on Niva.ID fields.

Parameters:

Name	Type	Description	Default
layer	qgis.core.QgsVectorLayer	IACS layer.	required

Returns:

Type	Description
qgis.core.QgsVectorLayer	Filtered IACS.

Source code in niva/core/prediction/data_manager.py

def filter_iacs(self, layer: QgsVectorLayer) -> QgsVectorLayer:
    """Filter IACS on Niva.ID fields.

    Args:
        layer (QgsVectorLayer): IACS layer.

    Returns:
        Filtered IACS.
    """
    filtered = layer.materialize(
        QgsFeatureRequest().setFilterExpression(f'"ID.NIVA" < 41')
    )
    return filtered

GridDataManager() dataclass

Bases: niva.core.prediction.data_manager.BaseDataManager

DataManager for grid based predictor.

compute(cell) abstractmethod

Compute a sample from cell geometry selection.

Parameters:

Name	Type	Description	Default
cell	qgis.core.QgsGeometry	Cell to compute metrics on.	required

Returns:

Type	Description
niva.core.sample.Sample	Sample of variables.

Source code in niva/core/prediction/data_manager.py

@abstractmethod
def compute(self, cell: QgsGeometry) -> Sample:
    """Compute a sample from cell geometry selection.

    Args:
        cell (QgsGeometry): Cell to compute metrics on.

    Returns:
        Sample of variables.
    """
    pass

crop_feat(feature, geometry)

Crop a feat with intersection geometry.

Parameters:

Name	Type	Description	Default
feature	qgis.core.QgsFeature	Fetaure to crop.	required
geometry	qgis.core.QgsGeometry	Geometry to compute intersection.	required

Returns:

Type	Description
typing.List[qgis.core.QgsFeature]	Cropped feature.

Source code in niva/core/prediction/data_manager.py

def crop_feat(self, feature: QgsFeature, geometry: QgsGeometry) -> List[QgsFeature]:
    """Crop a feat with intersection geometry.

    Args:
        feature (QgsFeature): Fetaure to crop.
        geometry (QgsGeometry): Geometry to compute intersection.

    Returns:
        Cropped feature.
    """
    feature.setGeometry(feature.geometry().intersection(geometry))
    return feature

get_features(layer, index, cell, crop=True)

Gather all features from a layer that intersect a geometry/cell.

Parameters:

Name	Type	Description	Default
layer	qgis.core.QgsVectorLayer	Layer to gather features from.	required
cell	qgis.core.QgsGeometry	Geometry to gather feats.	required
crop	bool, **optional**	If True, crop feats to geometry. Defaults to False.	True

Returns:

Type	Description
typing.List[qgis.core.QgsFeature]	List of geometries extracted.

Source code in niva/core/prediction/data_manager.py

def get_features(
    self,
    layer: QgsVectorLayer,
    index: QgsSpatialIndex,
    cell: QgsGeometry,
    crop=True,
) -> List[QgsFeature]:
    """Gather all features from a layer that intersect a geometry/cell.

    Args:
        layer (QgsVectorLayer): Layer to gather features from.
        cell (QgsGeometry): Geometry to gather feats.
        crop (bool, **optional**): If True, crop feats to geometry. Defaults to False.

    Returns:
        List of geometries extracted.
    """
    features_ids = index.intersects(cell.boundingBox())
    features: List[QgsFeature] = layer.getFeatures(features_ids)
    features: List[QgsFeature] = [
        f for f in features if f.geometry().intersects(cell)
    ]
    if crop:
        features = [self.crop_feat(f, cell) for f in features]
    return features

T1GridDataManager(*, iacs, artificial, naa=None) dataclass

Bases: niva.core.prediction.data_manager.GridDataManager

Data manager for Contra Tier 1 with grid based predictions.

Parameters:

Name	Type	Description	Default
iacs	qgis.core.QgsVectorLayer	Layer of IACS data. Warning: Provide IACS with Niva.ID field.	required
artificial	qgis.core.QgsVectorLayer	Layer of artificial data.	required
naa	qgis.core.QgsVectorLayer	Layer of NAA data.	None

Attributes:

Name	Type	Description
iacs	qgis.core.QgsVectorLayer	Layer of IACS data.
artificial	qgis.core.QgsVectorLayer	Layer of artificial data.
naa	Union[qgis.core.QgsVectorLayer, None]	Layer of NAA data or None if not passed to constructor.
iacs_index	qgis.core.QgsSpatialIndex	Spatial index for layer iacs.
artificial_index	qgis.core.QgsSpatialIndex	Spatial index for layer artificial.
naa_index	qgis.core.QgsSpatialIndex	Spatial index for layer naa.

compute(cell)

Compute metrics and build sample for cell.

Parameters:

Name	Type	Description	Default
cell	qgis.core.QgsGeometry	Cell geometry to compute metric on.	required

Returns:

Type	Description
niva.core.sample.T1Sample	Tier 1 sample for cell.

Source code in niva/core/prediction/data_manager.py

def compute(self, cell: QgsGeometry) -> T1Sample:
    """Compute metrics and build sample for cell.

    Args:
        cell (QgsGeometry): Cell geometry to compute metric on.

    Returns:
        Tier 1 sample for cell.
    """
    # Gather features
    iacs_feats = self.get_features(self.iacs, self.iacs_index, cell)
    artificial_feats = self.get_features(
        self.artificial, self.artificial_index, cell, crop=False
    )
    # Group geometries
    iacs = QgsGeometry.unaryUnion([f.geometry() for f in iacs_feats])
    artificial = QgsGeometry.unaryUnion([f.geometry() for f in artificial_feats])
    # get buildings with aoe geometry
    buildings = self._get_buildings(artificial_feats)
    # if naa apply geometries transformations.
    if self.naa:
        naa_feats = self.get_features(self.naa, self.naa_index, cell)
        naa_union = QgsGeometry.unaryUnion([f.geometry() for f in naa_feats])
        naa_union = naa_union.difference(artificial)
        iacs = iacs.difference(naa_union)
        buildings = buildings.difference(naa_union)
    else:
        naa_union = QgsGeometry.fromWkt("POLYGON EMPTY")

    artificial = QgsGeometry.unaryUnion([artificial, buildings])
    if not naa_union.isEmpty():
        artificial = artificial.difference(naa_union)
    artificial = artificial.intersection(cell)

    # compute metrics
    snc = semi_natural_cover(artificial, iacs, cell.area())
    mfs = mean_field_size(iacs_feats)
    rich, simpson = diversity(iacs_feats)

    return T1Sample(richness=rich, simpson=simpson, snc=snc, mfs=mfs)

Predictor(context=QgsProcessingContext(), feedback=QgsProcessingFeedback(), *args, **kwargs)

Bases: niva.core.mixins.QgsLogicMixin

Base predictor class.

Source code in niva/core/mixins.py

def __init__(
    self,
    context: QgsProcessingContext = QgsProcessingContext(),
    feedback: QgsProcessingFeedback = QgsProcessingFeedback(),
    *args,
    **kwargs,
):
    """
    Args:
        context (QgsProcessingContext, **optional**): Context to use within the class.
        feedback (QgsProcessingFeedback, **optional**): Feedback to use within the class.
    """
    self.context = context
    self.feedback = feedback
    super().__init__(*args, **kwargs)

run() abstractmethod

Run the prediction and output a QgsVectorLayer with predictions.

Source code in niva/core/prediction/predictor.py

@abstractmethod
def run(self) -> QgsVectorLayer:
    """Run the prediction and output a QgsVectorLayer with predictions."""
    pass

GridPredictor(tree, data_manager, grid, *args, **kwargs)

Bases: niva.core.prediction.predictor.Predictor

Base predictor class for grid predictions.

Attributes:

Name	Type	Description
tree	niva.core.contra.ContraTree	ContratTree for prediction.
data_manager	niva.core.prediction.data_manager.BaseDataManager	Data manager.
grid	qgis.core.QgsVectorLayer	Grdi to predict on.

Parameters:

Name	Type	Description	Default
tree	niva.core.contra.ContraTree	ContratTree for prediction.	required
data_manager	niva.core.prediction.data_manager.BaseDataManager	Data manager.	required
grid	qgis.core.QgsVectorLayer	Grdi to predict on.	required

Source code in niva/core/prediction/predictor.py

def __init__(
    self,
    tree: ContraTree,
    data_manager: BaseDataManager,
    grid: QgsVectorLayer,
    *args,
    **kwargs,
):
    """
    Args:
        tree (ContraTree): ContratTree for prediction.
        data_manager (BaseDataManager): Data manager.
        grid (QgsVectorLayer): Grdi to predict on.
    """

    self.data_manager = data_manager
    self.tree = tree
    self.grid = grid
    super().__init__(*args, **kwargs)

run()

Run the predictions by iterating over each grid cells.

Returns:

Type	Description
qgis.core.QgsVectorLayer	Prediction layer.

Source code in niva/core/prediction/predictor.py

def run(self) -> QgsVectorLayer:
    """Run the predictions by iterating over each grid cells.

    Returns:
        Prediction layer.
    """
    predictions = []
    total = self.grid.featureCount()
    for i, cell in enumerate(self.grid.getFeatures()):
        sample = self.data_manager.compute(cell.geometry())
        sample.mdi, sample.novelty, sample.branch = self.tree.predict(sample)
        pred_feat = sample.to_feature()
        pred_feat.setGeometry(cell.geometry())
        predictions.append(pred_feat)
        self.feedback.setProgress((i / total) * 100)
        self.cancel()

    output = QgsVectorLayer(
        f"Polygon?crs={self.grid.crs().authid()}", "predictions", "memory"
    )
    output.dataProvider().addAttributes(list(predictions[0].fields()))
    output.updateFields()
    output.dataProvider().addFeatures(predictions)
    return output

NivaMdi

Base class for Niva's prediction.

predict_mdi() abstractmethod

Predict Niva MultiDiversityIndex.

Source code in niva/core/prediction/predictor.py

@abstractmethod
def predict_mdi(self):
    """Predict Niva MultiDiversityIndex."""
    pass

NivaMdiGridT1(grid, iacs, artificial, naa=None, tree_name='T1', *args, **kwargs)

Bases: niva.core.prediction.predictor.NivaMdi, niva.core.mixins.QgsLogicMixin

Manage Niva prediction for Tier 1 on a grid.

Parameters:

Name	Type	Description	Default
grid	qgis.core.QgsVectorLayer	Grid layer to predict on.	required
iacs	qgis.core.QgsVectorLayer	IACS layer with Niva.ID field.	required
artificial	qgis.core.QgsVectorLayer	Artificial layer.	required
naa	Union[QgsVectorLayer, None], **optional**	NAA layer. Defaults to None.	None
tree_name	str, **optional**	Name of tree config into ContraTreeFactory. Defaults to "T1".	'T1'

Attributes:

Name	Type	Description
tree	niva.core.contra.ContraTree	Contra decision tree to use for prediction.
data_manager	niva.core.prediction.data_manager.T1GridDataManager	Data manager for grid T1.
predictor	niva.core.prediction.predictor.GridPredictor	Predictor to use for T1 grid prediction.

Source code in niva/core/prediction/predictor.py

def __init__(
    self,
    grid: QgsVectorLayer,
    iacs: QgsVectorLayer,
    artificial: QgsVectorLayer,
    naa: Union[QgsVectorLayer, None] = None,
    tree_name: str = "T1",
    *args,
    **kwargs,
):
    """
    Args:
        grid (QgsVectorLayer): Grid layer to predict on.
        iacs (QgsVectorLayer): IACS layer with Niva.ID field.
        artificial (QgsVectorLayer): Artificial layer.
        naa (Union[QgsVectorLayer, None], **optional**): NAA layer. Defaults to None.
        tree_name (str, **optional**): Name of tree config into ContraTreeFactory. Defaults to "T1".

    Attributes:
        tree (ContraTree): Contra decision tree to use for prediction.
        data_manager (T1GridDataManager): Data manager for grid T1.
        predictor (GridPredictor): Predictor to use for T1 grid prediction.
    """
    self.tree = ContraTreeFactory.build(tree_name)
    layers = [artificial, iacs, grid]
    if naa:
        layers.append(naa)
    self.data_manager = T1GridDataManager(iacs=iacs, artificial=artificial, naa=naa)
    self.predictor = GridPredictor(
        tree=self.tree, data_manager=self.data_manager, grid=grid, *args, **kwargs
    )

predict_mdi()

Run prediction.

Returns:

Type	Description
qgis.core.QgsVectorLayer	Prediction layer.

Source code in niva/core/prediction/predictor.py

def predict_mdi(self) -> QgsVectorLayer:
    """Run prediction.

    Returns:
        Prediction layer.
    """
    return self.predictor.run()