Tree

• Tree
  • Tree
  • Tree
    • root
    • plot
    • pretty
  • Conditions
  • AbstractCondition
    • missing
    • score
    • pretty
  • NumericalHigherThanCondition
    • attribute
    • missing
    • score
    • threshold
    • pretty
  • CategoricalIsInCondition
    • attribute
    • mask
    • missing
    • score
    • pretty
  • CategoricalSetContainsCondition
    • attribute
    • mask
    • missing
    • score
    • pretty
  • DiscretizedNumericalHigherThanCondition
    • attribute
    • missing
    • score
    • threshold_idx
    • pretty
  • IsMissingInCondition
    • attribute
    • missing
    • score
    • pretty
  • IsTrueCondition
    • attribute
    • missing
    • score
    • pretty
  • NumericalSparseObliqueCondition
    • attributes
    • missing
    • score
    • threshold
    • weights
    • pretty
  • Nodes
  • AbstractNode
    • is_leaf
    • pretty
  • Leaf
    • is_leaf
    • value
    • pretty
  • NonLeaf
    • condition
    • is_leaf
    • neg_child
    • pos_child
    • value
    • pretty
  • Values
  • AbstractValue
    • num_examples
    • pretty
  • RegressionValue
    • num_examples
    • standard_deviation
    • value
    • pretty
  • ProbabilityValue
    • num_examples
    • probability
    • pretty
  • UpliftValue
    • num_examples
    • treatment_effect
    • pretty

Tree

Tree dataclass

Tree(root: AbstractNode)

root instance-attribute

root: AbstractNode

plot

plot(
    dataspec: DataSpecification,
    max_depth: Optional[int],
    label_classes: Optional[Sequence[str]],
    options: Optional[PlotOptions] = None,
    d3js_url: str = "https://d3js.org/d3.v6.min.js",
) -> TreePlot

Plots a decision tree.

Parameters:

Name	Type	Description	Default
dataspec	DataSpecification	Dataspec of the tree.	required
max_depth	Optional[int]	Maximum tree depth of the plot. Set to None for full depth.	required
label_classes	Optional[Sequence[str]]	For classification, label classes of the dataset.	required
options	Optional[PlotOptions]	Advanced options for plotting. Set to None for default style.	None
d3js_url	str	URL to load the d3.js library from.	'https://d3js.org/d3.v6.min.js'

Returns:

Type	Description
TreePlot	The html content displaying the tree.

pretty

pretty(
    dataspec: DataSpecification,
    max_depth: Optional[int] = 6,
) -> str

Returns a printable representation of the decision tree.

Usage example:

model = ydf.load_model("my_model")
tree = model.get_tree(0)
print(tree.pretty(model.data_spec()))

Parameters:

Name	Type	Description	Default
dataspec	DataSpecification	Dataspec of the tree.	required
max_depth	Optional[int]	Maximum printed depth.	6

Conditions

AbstractCondition dataclass

AbstractCondition(missing: bool, score: float)

Generic condition.

Attrs

missing: Result of the evaluation of the condition if the input feature is missing. score: Score of a condition. The semantic depends on the learning algorithm.

missing instance-attribute

missing: bool

score instance-attribute

score: float

pretty abstractmethod

pretty(dataspec: DataSpecification) -> str

NumericalHigherThanCondition dataclass

NumericalHigherThanCondition(
    missing: bool,
    score: float,
    attribute: int,
    threshold: float,
)

Bases: AbstractCondition

Condition of the form "attribute >= threshold".

Attrs

attribute: Attribute tested by the condition. threshold: Threshold.

attribute instance-attribute

attribute: int

missing instance-attribute

missing: bool

score instance-attribute

score: float

threshold instance-attribute

threshold: float

pretty

pretty(dataspec: DataSpecification) -> str

CategoricalIsInCondition dataclass

CategoricalIsInCondition(
    missing: bool,
    score: float,
    attribute: int,
    mask: Sequence[int],
)

Bases: AbstractCondition

Condition of the form "attribute in mask".

Attrs

attribute: Attribute tested by the condition. mask: Sorted mask values.

attribute instance-attribute

attribute: int

mask instance-attribute

mask: Sequence[int]

missing instance-attribute

missing: bool

score instance-attribute

score: float

pretty

pretty(dataspec: DataSpecification) -> str

CategoricalSetContainsCondition dataclass

CategoricalSetContainsCondition(
    missing: bool,
    score: float,
    attribute: int,
    mask: Sequence[int],
)

Bases: AbstractCondition

Condition of the form "attribute intersect mask != empty".

Attrs

attribute: Attribute tested by the condition. mask: Sorted mask values.

attribute instance-attribute

attribute: int

mask instance-attribute

mask: Sequence[int]

missing instance-attribute

missing: bool

score instance-attribute

score: float

pretty

pretty(dataspec: DataSpecification) -> str

DiscretizedNumericalHigherThanCondition dataclass

DiscretizedNumericalHigherThanCondition(
    missing: bool,
    score: float,
    attribute: int,
    threshold_idx: int,
)

Bases: AbstractCondition

Condition of the form "attribute >= bounds[threshold]".

Attrs

attribute: Attribute tested by the condition. threshold_idx: Index of threshold in dataspec.

attribute instance-attribute

attribute: int

missing instance-attribute

missing: bool

score instance-attribute

score: float

threshold_idx instance-attribute

threshold_idx: int

pretty

pretty(dataspec: DataSpecification) -> str

IsMissingInCondition dataclass

IsMissingInCondition(
    missing: bool, score: float, attribute: int
)

Bases: AbstractCondition

Condition of the form "attribute is missing".

Attrs

attribute: Attribute (or one of the attributes) tested by the condition.

attribute instance-attribute

attribute: int

missing instance-attribute

missing: bool

score instance-attribute

score: float

pretty

pretty(dataspec: DataSpecification) -> str

IsTrueCondition dataclass

IsTrueCondition(
    missing: bool, score: float, attribute: int
)

Bases: AbstractCondition

Condition of the form "attribute is true".

Attrs

attribute: Attribute tested by the condition.

attribute instance-attribute

attribute: int

missing instance-attribute

missing: bool

score instance-attribute

score: float

pretty

pretty(dataspec: DataSpecification) -> str

NumericalSparseObliqueCondition dataclass

NumericalSparseObliqueCondition(
    missing: bool,
    score: float,
    attributes: Sequence[int],
    weights: Sequence[float],
    threshold: float,
)

Bases: AbstractCondition

Condition of the form "attributes * weights >= threshold".

Attrs

attributes: Attribute tested by the condition. weights: Weights for each of the attributes. threshold: Threshold value of the condition.

attributes instance-attribute

attributes: Sequence[int]

missing instance-attribute

missing: bool

score instance-attribute

score: float

threshold instance-attribute

threshold: float

weights instance-attribute

weights: Sequence[float]

pretty

pretty(dataspec: DataSpecification) -> str

Nodes

AbstractNode

is_leaf abstractmethod property

is_leaf: bool

Tells if a node is a leaf.

pretty abstractmethod

pretty(
    dataspec: DataSpecification,
    prefix: str,
    is_pos: Optional[bool],
    depth: int,
    max_depth: Optional[int],
) -> str

Leaf dataclass

Leaf(value: AbstractValue)

Bases: AbstractNode

is_leaf property

is_leaf: bool

Tells if a node is a leaf.

value instance-attribute

value: AbstractValue

pretty

pretty(
    dataspec: DataSpecification,
    prefix: str,
    is_pos: Optional[bool],
    depth: int,
    max_depth: Optional[int],
) -> str

NonLeaf dataclass

NonLeaf(
    value: Optional[AbstractValue] = None,
    condition: Optional[AbstractCondition] = None,
    pos_child: Optional[AbstractNode] = None,
    neg_child: Optional[AbstractNode] = None,
)

Bases: AbstractNode

condition class-attribute instance-attribute

condition: Optional[AbstractCondition] = None

is_leaf property

is_leaf: bool

Tells if a node is a leaf.

neg_child class-attribute instance-attribute

neg_child: Optional[AbstractNode] = None

pos_child class-attribute instance-attribute

pos_child: Optional[AbstractNode] = None

value class-attribute instance-attribute

value: Optional[AbstractValue] = None

pretty

pretty(
    dataspec: DataSpecification,
    prefix: str,
    is_pos: Optional[bool],
    depth: int,
    max_depth: Optional[int],
) -> str

Values

AbstractValue dataclass

AbstractValue(num_examples: float)

A generic value/prediction/output.

Attrs

num_examples: Number of examples in the node with weight.

num_examples instance-attribute

num_examples: float

pretty abstractmethod

pretty() -> str

RegressionValue dataclass

RegressionValue(
    num_examples: float,
    value: float,
    standard_deviation: Optional[float] = None,
)

Bases: AbstractValue

The regression value of a regressive tree.

Can also be used in gradient-boosted-trees for classification and ranking.

Attrs

value: Value of the tree. The semantic depends on the tree: For Regression Random Forest and Regression GBDT, this value is a regressive value in the same unit as the label. For classification and ranking GBDTs, this value is a logit. standard_deviation: Optional standard deviation attached to the value.

num_examples instance-attribute

num_examples: float

standard_deviation class-attribute instance-attribute

standard_deviation: Optional[float] = None

value instance-attribute

value: float

pretty

pretty() -> str

ProbabilityValue dataclass

ProbabilityValue(
    num_examples: float, probability: Sequence[float]
)

Bases: AbstractValue

A probability distribution value.

Used for random Forest / CART classification trees.

Attrs

probability: An array of probabilities of the label classes i.e. the i-th value is the probability of the "label_value_idx_to_value(..., i)" class. Note that the first value is reserved for the Out-of-vocabulary

num_examples instance-attribute

num_examples: float

probability instance-attribute

probability: Sequence[float]

pretty

pretty() -> str

UpliftValue dataclass

UpliftValue(
    num_examples: float, treatment_effect: Sequence[float]
)

Bases: AbstractValue

The uplift value of a classification or regression uplift tree.

Attrs

treatment_effect: An array of the effects on the treatment groups. The i-th element of this array is the effect of the "i+1"th treatment compared to the control group.

num_examples instance-attribute

num_examples: float

treatment_effect instance-attribute

treatment_effect: Sequence[float]

pretty

pretty() -> str